Contributors should read our contributors guide for instructions on how to update the changelog.
This document contains a historical list of changes between releases. Only changes that impact end-user behavior are listed; changes to documentation or internal API changes are not present.
-
A new
discovery.process
component for discovering Linux OS processes on the current host. (@korniltsev) -
A new
pyroscope.java
component for profiling Java processes using async-profiler. (@korniltsev)
-
Include line numbers in profiles produced by
pyrsocope.java
component. (@korniltsev) -
Add an option to the windows static mode installer for expanding environment vars in the yaml config. (@erikbaranowski)
-
Add authentication support to
loki.source.awsfirehose
(@sberz) -
Sort kubelet endpoint to reduce pressure on K8s's API server and watcher endpoints. (@hainenber)
-
Expose
physical_disk
collector fromwindows_exporter
v0.24.0 to Flow configuration. (@hainenber)
-
Fix an issue in
remote.s3
where the exported content of an object would be an empty string ifremote.s3
failed to fully retrieve the file in a single read call. (@grafana/agent-squad) -
Utilize the
instance
Argument ofprometheus.exporter.kafka
when set. (@akhmatov-s) -
Fix a duplicate metrics registration panic when sending metrics to an static mode metric instance's write handler. (@tpaschalis)
-
Fix issue causing duplicate logs when a docker target is restarted. (@captncraig)
-
Fix an issue where blocks having the same type and the same label across modules could result in missed updates. (@thampiotr)
-
Removed support for Windows 2012 in line with Microsoft end of life. (@mattdurham)
-
Split instance ID and component groupings into separate panels for
remote write active series by component
in the Flow mixin. (@tristanburgess) -
Updated dependency to add support for Go 1.22 (@stefanb)
- Fixes following vulnerabilities (@hainenber)
- Fix issue where installing the Windows Agent Flow installer would hang then crash. (@mattdurham)
-
otelcol.receiver.prometheus
will drop allotel_scope_info
metrics when converting them to OTLP. (@wildum)- If the
otel_scope_info
metric has labelsotel_scope_name
andotel_scope_version
, their values will be used to set OTLP Instrumentation Scope name and version respectively. - Labels of
otel_scope_info
metrics other thanotel_scope_name
andotel_scope_version
are added as scope attributes with the matching name and version.
- If the
-
The
target
block inprometheus.exporter.blackbox
requires a mandatoryname
argument instead of a block label. (@hainenber) -
In the azure exporter, dimension options will no longer be validated by the Azure API. (@kgeckhart)
- This change will not break any existing configurations and you can opt in to validation via the
validate_dimensions
configuration option. - Before this change, pulling metrics for azure resources with variable dimensions required one configuration per metric + dimension combination to avoid an error.
- After this change, you can include all metrics and dimensions in a single configuration and the Azure APIs will only return dimensions which are valid for the various metrics.
- This change will not break any existing configurations and you can opt in to validation via the
-
A new
discovery.ovhcloud
component for discovering scrape targets on OVHcloud. (@ptodev) -
Allow specifying additional containers to run. (@juangom)
-
Flow Windows service: Support environment variables. (@jkroepke)
-
Allow disabling collection of root Cgroup stats in
prometheus.exporter.cadvisor
(flow mode) and thecadvisor
integration (static mode). (@hainenber) -
Grafana Agent on Windows now automatically restarts on failure. (@hainenber)
-
Added metrics, alerts and dashboard visualisations to help diagnose issues with unhealthy components and components that take too long to evaluate. (@thampiotr)
-
The
http
config block may now reference exports from any component. Previously, onlyremote.*
andlocal.*
components could be referenced without a circular dependency. (@rfratto) -
Add support for Basic Auth-secured connection with Elasticsearch cluster using
prometheus.exporter.elasticsearch
. (@hainenber) -
Add a
resource_to_telemetry_conversion
argument tootelcol.exporter.prometheus
for converting resource attributes to Prometheus labels. (@hainenber) -
pyroscope.ebpf
support python on arm64 platforms. (@korniltsev) -
otelcol.receiver.prometheus
does not drop histograms without buckets anymore. (@wildum) -
Added exemplars support to
otelcol.receiver.prometheus
. (@wildum) -
mimir.rules.kubernetes
may now retry its startup on failure. (@hainenber) -
Added links between compatible components in the documentation to make it easier to discover them. (@thampiotr)
-
Allow defining
HTTPClientConfig
fordiscovery.ec2
. (@cmbrad) -
The
remote.http
component can optionally define a request body. (@tpaschalis) -
Added support for
loki.write
to flush WAL on agent shutdown. (@thepalbi) -
Add support for
integrations-next
static to flow config conversion. (@erikbaranowski) -
Add support for passing extra arguments to the static converter such as
-config.expand-env
. (@erikbaranowski) -
Added 'country' mmdb-type to log pipeline-stage geoip. (@superstes)
-
Azure exporter enhancements for flow and static mode, (@kgeckhart)
- Allows for pulling metrics at the Azure subscription level instead of resource by resource
- Disable dimension validation by default to reduce the number of exporter instances needed for full dimension coverage
-
Add
max_cache_size
toprometheus.relabel
to allow configurability instead of hard coded 100,000. (@mattdurham) -
Add support for
http_sd_config
within ascrape_config
for prometheus to flow config conversion. (@erikbaranowski) -
discovery.lightsail
now supports additional parameters for configuring HTTP client settings. (@ptodev) -
Add
sample_age_limit
to remote_write config to drop samples older than a specified duration. (@marctc) -
Handle paths in the Kubelet URL for
discovery.kubelet
. (@petewall) -
loki.source.docker
now deduplicates targets which report the same container ID. (@tpaschalis)
-
Update
pyroscope.ebpf
to fix a logical bug causing to profile to many kthreads instead of regular processes grafana/pyroscope#2778 (@korniltsev) -
Update
pyroscope.ebpf
to produce more optimal pprof profiles for python processes grafana/pyroscope#2788 (@korniltsev) -
In Static mode's
traces
subsystem,spanmetrics
used to be generated prior to load balancing. This could lead to inaccurate metrics. This issue only affects Agents using bothspanmetrics
andload_balancing
, when running in a load balanced cluster with more than one Agent instance. (@ptodev) -
Fixes
loki.source.docker
a behavior that synced an incomplete list of targets to the tailer manager. (@FerdinandvHagen) -
Fixes
otelcol.connector.servicegraph
store ttl default value from 2ms to 2s. (@rlankfo) -
Add staleness tracking to labelstore to reduce memory usage. (@mattdurham)
-
Fix issue where
prometheus.exporter.kafka
would crash when configuringsasl_password
. (@rfratto) -
Fix performance issue where perf lib where clause was not being set, leading to timeouts in collecting metrics for windows_exporter. (@mattdurham)
-
Fix nil panic when using the process collector with the windows exporter. (@mattdurham)
-
Bump github.com/IBM/sarama from v1.41.2 to v1.42.1
-
Attach unique Agent ID header to remote-write requests. (@captncraig)
-
Update to v2.48.1 of
github.com/prometheus/prometheus
. Previously, a custom fork of v2.47.2 was used. The custom fork of v2.47.2 also contained prometheus#12729 and prometheus#12677.
- Fix CVE-2023-47108 by updating
otelgrpc
from v0.45.0 to v0.46.0. (@hainenber)
- Agent Management: Introduce support for templated configuration. (@jcreixell)
-
Permit
X-Faro-Session-ID
header in CORS requests for thefaro.receiver
component (flow mode) and theapp_agent_receiver
integration (static mode). (@cedricziel) -
Fix issue with windows_exporter defaults not being set correctly. (@mattdurham)
-
Fix agent crash when process null OTel's fan out consumers. (@hainenber)
-
Fix issue in
prometheus.operator.*
where targets would be dropped if two crds share a common prefix in their names. (@Paul424, @captncraig) -
Fix issue where
convert
command would generate incorrect Flow Mode config when providedpromtail
configuration that usesdocker_sd_configs
(@thampiotr) -
Fix converter issue with
loki.relabel
andmax_cache_size
being set to 0 instead of default (10_000). (@mattdurham)
- Add Agent Deploy Mode to usage report. (@captncraig)
-
Remove
otelcol.exporter.jaeger
component (@hainenber) -
In the mysqld exporter integration, some metrics are removed and others are renamed. (@marctc)
- Removed metrics:
- "mysql_last_scrape_failed" (gauge)
- "mysql_exporter_scrapes_total" (counter)
- "mysql_exporter_scrape_errors_total" (counter)
- Metric names in the
info_schema.processlist
collector have been changed. - Metric names in the
info_schema.replica_host
collector have been changed. - Changes related to
replication_group_member_stats collector
:- metric "transaction_in_queue" was Counter instead of Gauge
- renamed 3 metrics starting with
mysql_perf_schema_transaction_
to start withmysql_perf_schema_transactions_
to be consistent with column names. - exposing only server's own stats by matching
MEMBER_ID
with@@server_uuid
resulting "member_id" label to be dropped.
- Removed metrics:
-
Added a new
stage.decolorize
stage toloki.process
component which allows to strip ANSI color codes from the log lines. (@thampiotr) -
Added a new
stage.sampling
stage toloki.process
component which allows to only process a fraction of logs and drop the rest. (@thampiotr) -
Added a new
stage.eventlogmessage
stage toloki.process
component which allows to extract data from Windows Event Log. (@thampiotr) -
Update version of River:
-
River now supports raw strings, which are strings surrounded by backticks instead of double quotes. Raw strings can span multiple lines, and do not support any escape sequences. (@erikbaranowski)
-
River now permits using
[]
to access non-existent keys in an object. When this is done, the access evaluates tonull
, such that{}["foo"] == null
is true. (@rfratto)
-
-
Added support for python profiling to
pyroscope.ebpf
component. (@korniltsev) -
Added support for native Prometheus histograms to
otelcol.exporter.prometheus
(@wildum) -
Windows Flow Installer: Add /CONFIG /DISABLEPROFILING and /DISABLEREPORTING flag (@jkroepke)
-
Add queueing logs remote write client for
loki.write
when WAL is enabled. (@thepalbi) -
New Grafana Agent Flow components:
otelcol.processor.filter
- filters OTLP telemetry data using OpenTelemetry Transformation Language (OTTL). (@hainenber)otelcol.receiver.vcenter
- receives metrics telemetry data from vCenter. (@marctc)
-
Agent Management: Introduce support for remotely managed external labels for logs. (@jcreixell)
-
The
loki.write
WAL now has snappy compression enabled by default. (@thepalbi) -
Allow converting labels to structured metadata with Loki's structured_metadata stage. (@gonzalesraul)
-
Improved performance of
pyroscope.scrape
component when working with a large number of targets. (@cyriltovena) -
Added support for comma-separated list of fields in
source
option and a newseparator
option indrop
stage ofloki.process
. (@thampiotr) -
The
loki.source.docker
component now allows connecting to Docker daemons over HTTP(S) and setting up TLS credentials. (@tpaschalis) -
Added an
exclude_event_message
option toloki.source.windowsevent
in flow mode, which excludes the human-friendly event message from Windows event logs. (@ptodev) -
Improve detection of rolled log files in
loki.source.kubernetes
andloki.source.podlogs
(@slim-bean). -
Support clustering in
loki.source.kubernetes
(@slim-bean). -
Support clustering in
loki.source.podlogs
(@rfratto). -
Make component list sortable in web UI. (@hainenber)
-
Adds new metrics (
mssql_server_total_memory_bytes
,mssql_server_target_memory_bytes
, andmssql_available_commit_memory_bytes
) formssql
integration (@StefanKurek). -
Grafana Agent Operator:
config-reloader
container no longer runs as root. (@rootmout) -
Added support for replaying not sent data for
loki.write
when WAL is enabled. (@thepalbi) -
Make the result of 'discovery.kubelet' support pods that without ports, such as k8s control plane static pods. (@masonmei)
-
Added support for unicode strings in
pyroscope.ebpf
python profiles. (@korniltsev) -
Improved resilience of graph evaluation in presence of slow components. (@thampiotr)
-
Updated windows exporter to use prometheus-community/windows_exporter commit 1836cd1. (@mattdurham)
-
Allow agent to start with
module.git
config if cached before. (@hainenber) -
Adds new optional config parameter
query_config
tomssql
integration to allow for custom metrics (@StefanKurek)
-
Set exit code 1 on grafana-agentctl non-runnable command. (@fgouteroux)
-
Fixed an issue where
loki.process
validation for stagemetric.counter
was allowing invalid combination of configuration options. (@thampiotr) -
Fixed issue where adding a module after initial start, that failed to load then subsequently resolving the issue would cause the module to permanently fail to load with
id already exists
error. (@mattdurham) -
Allow the usage of encodings other than UTF8 to be used with environment variable expansion. (@mattdurham)
-
Fixed an issue where native histogram time series were being dropped silently. (@krajorama)
-
Fix validation issue with ServiceMonitors when scrape timeout is greater than interval. (@captncraig)
-
Static mode's spanmetrics processor will now prune histograms when the dimension cache is pruned. Dimension cache was always pruned but histograms were not being pruned. This caused metric series created by the spanmetrics processor to grow unbounded. Only static mode has this issue. Flow mode's
otelcol.connector.spanmetrics
does not have this bug. (@nijave) -
Prevent logging errors on normal shutdown in
loki.source.journal
. (@wildum) -
Break on iterate journal failure in
loki.source.journal
. (@wildum) -
Fix file descriptor leak in
loki.source.journal
. (@wildum) -
Fixed a bug in River where passing a non-string key to an object (such as
{}[true]
) would incorrectly report that a number type was expected instead. (@rfratto) -
Include Faro Measurement
type
field infaro.receiver
Flow component and legacyapp_agent_receiver
integration. (@rlankfo) -
Mark
password
argument ofloki.source.kafka
as asecret
rather than astring
. (@harsiddhdave44) -
Fixed a bug where UDP syslog messages were never processed (@joshuapare)
-
Updating configuration for
loki.write
no longer drops data. (@thepalbi) -
Fixed a bug in WAL where exemplars were recorded before the first native histogram samples for new series, resulting in remote write sending the exemplar first and Prometheus failing to ingest it due to missing series. (@krajorama)
-
Fixed an issue in the static config converter where exporter instance values were not being mapped when translating to flow. (@erikbaranowski)
-
Fix a bug which prevented Agent from running
otelcol.exporter.loadbalancing
with arouting_key
oftraceID
. (@ptodev) -
Added Kubernetes service resolver to static node's loadbalancing exporter and to Flow's
otelcol.exporter.loadbalancing
. (@ptodev) -
Fix default configuration file
grafana-agent-flow.river
used in downstream packages. (@bricewge) -
Fix converter output for prometheus.exporter.windows to not unnecessarily add empty blocks. (@erikbaranowski)
-
Bump
mysqld_exporter
version to v0.15.0. (@marctc) -
Bump
github-exporter
version to 1.0.6. (@marctc) -
Use Go 1.21.4 for builds. (@rfratto)
-
Change User-Agent header for outbound requests to include agent-mode, goos, and deployment mode. Example
GrafanaAgent/v0.38.0 (flow; linux; docker)
(@captncraig) -
loki.source.windowsevent
andloki.source.*
changed to use a more robust positions file to prevent corruption on reboots when writing the positions file. (@mattdurham)
- Added an
add_metric_suffixes
option tootelcol.exporter.prometheus
in flow mode, which configures whether to add type and unit suffixes to metrics names. (@mar4uk)
-
Fix a bug where reloading the configuration of a
loki.write
component lead to a panic. (@tpaschalis) -
Added Kubernetes service resolver to static node's loadbalancing exporter and to Flow's
otelcol.exporter.loadbalancing
. (@ptodev)
-
Fixed an issue where native histogram time series were being dropped silently. (@krajorama)
-
Fix an issue where
remote.vault
ignored thenamespace
argument. (@rfratto) -
Fix an issue with static mode and
promtail
converters, where static targets did not correctly default tolocalhost
when not provided. (@thampiotr) -
Fixed some converter diagnostics so they show as warnings rather than errors. Improve clarity for various diagnostics. (@erikbaranowski)
-
Wire up the agent exporter integration for the static converter. (@erikbaranowski)
- Upgrade OpenTelemetry Collector packages to version 0.87 (@ptodev):
otelcol.receiver.kafka
has a newheader_extraction
block to extract headers from Kafka records.otelcol.receiver.kafka
has a newversion
argument to change the version of the SASL Protocol for SASL authentication.
-
Fix the handling of the
--cluster.join-addresses
flag causing an invalid comparison with the mutually-exclusive--cluster.discover-peers
. (@tpaschalis) -
Fix an issue with the static to flow converter for blackbox exporter modules config not being included in the river output. (@erikbaranowski)
-
Fix issue with default values in
discovery.nomad
. (@marctc)
-
Update Prometheus dependency to v2.47.2. (@tpaschalis)
-
Allow Out of Order writing to the WAL for metrics. (@mattdurham)
-
Added new config options to spanmetrics processor in static mode (@ptodev):
aggregation_temporality
: configures whether to reset the metrics after flushing.metrics_flush_interval
: configures how often to flush generated metrics.
- Use Go 1.21.3 for builds. (@tpaschalis)
- Fix the initialization of the default namespaces map for the operator and the loki.source.kubernetes component. (@wildum)
-
Set
retry_on_http_429
totrue
by default in thequeue_config
block in static mode'sremote_write
. (@wildum) -
Renamed
non_indexed_labels
Loki processing stage tostructured_metadata
. (@vlad-diachenko) -
Include
otel_scope_name
andotel_scope_version
in all metrics forotelcol.exporter.prometheus
by default using a new argumentinclude_scope_labels
. (@erikbaranowski) -
Static mode Windows Certificate Filter no longer restricted to TLS 1.2 and specific cipher suites. (@mattdurham)
-
The
__meta_agent_integration*
and__meta_agent_hostname
labels have been removed from the targets exposed byprometheus.exporter.*
components and got replaced by the pair of__meta_component_name
and__meta_component_id
labels. (@tpaschalis) -
Flow: Allow
prometheus.exporter.unix
to be specified multiple times and used in modules. This now means allprometheus.exporter.unix
references will need a labelprometheus.exporter.unix "example"
. (@mattdurham)
-
New Grafana Agent Flow components:
discovery.consulagent
discovers scrape targets from Consul Agent. (@wildum)discovery.dockerswarm
discovers scrape targets from Docker Swarm. (@wildum)discovery.ionos
discovers scrape targets from the IONOS Cloud API. (@wildum)discovery.kuma
discovers scrape targets from the Kuma control plane. (@tpaschalis)discovery.linode
discovers scrape targets from the Linode API. (@captncraig)discovery.marathon
discovers scrape targets from Marathon servers. (@wildum)discovery.nerve
discovers scrape targets from AirBnB's Nerve. (@tpaschalis)discovery.scaleway
discovers scrape targets from Scaleway virtual instances and bare-metal machines. (@rfratto)discovery.serverset
discovers Serversets stored in Zookeeper. (@thampiotr)discovery.triton
discovers scrape targets from Triton Container Monitor. (@erikbaranowski)faro.receiver
accepts Grafana Faro-formatted telemetry data over the network and forwards it to other components. (@megumish, @rfratto)otelcol.connector.servicegraph
creates service graph metrics from spans. It is the flow mode equivalent to static mode'sservice_graphs
processor. (@ptodev)otelcol.connector.spanlogs
creates logs from spans. It is the flow mode equivalent to static mode'sautomatic_logging
processor. (@ptodev)otelcol.processor.k8sattributes
adds Kubernetes metadata as resource attributes to spans, logs, and metrics. (@acr92)otelcol.processor.probabilistic_sampler
samples logs and traces based on configuration options. (@mar4uk)otelcol.processor.transform
transforms OTLP telemetry data using the OpenTelemetry Transformation Language (OTTL). It is most commonly used for transformations on attributes.prometheus.exporter.agent
exposes the agent's internal metrics. (@hainenber)prometheus.exporter.azure
collects metrics from Azure. (@wildum)prometheus.exporter.cadvisor
exposes cAdvisor metrics. (@tpaschalis)prometheus.exporter.vsphere
exposes vmware vsphere metrics. (@marctc)remote.kubernetes.configmap
loads a configmap's data for use in other components (@captncraig)remote.kubernetes.secret
loads a secret's data for use in other components (@captncraig)
-
Flow: allow the HTTP server to be configured with TLS in the config file using the new
http
config block. (@rfratto) -
Clustering: add new flag
--cluster.max-join-peers
to limit the number of peers the system joins. (@wildum) -
Clustering: add a new flag
--cluster.name
to prevent nodes without this identifier from joining the cluster. (@wildum) -
Clustering: add IPv6 support when using advertise interfaces to assign IP addresses. (@wildum)
-
Add a
file_watch
block inloki.source.file
to configure how often to poll files from disk for changes viamin_poll_frequency
andmax_poll_frequency
. In static mode it can be configured in the globalfile_watch_config
viamin_poll_frequency
andmax_poll_frequency
. (@wildum) -
Flow: In
prometheus.exporter.blackbox
, allow setting labels for individual targets. (@spartan0x117) -
Add optional
nil_to_zero
config flag forYACE
which can be set in thestatic
,discovery
, ormetric
config blocks. (@berler) -
The
cri
stage inloki.process
can now be configured to limit line size. -
Flow: Allow
grafana-agent run
to accept a path to a directory of*.river
files. This will load all River files in the directory as a single configuration; component names must be unique across all loaded files. (@rfratto, @hainenber) -
Added support for
static
configuration conversion ingrafana-agent convert
andgrafana-agent run
commands. (@erikbaranowski) -
Flow: the
prometheus.scrape
component can now configure the scraping of Prometheus native histograms. (@tpaschalis) -
Flow: the
prometheus.remote_write
component now supports SigV4 and AzureAD authentication. (@ptodev)
-
Clustering: allow advertise interfaces to be configurable, with the possibility to select all available interfaces. (@wildum)
-
Deleted series will now be removed from the WAL sooner, allowing Prometheus remote_write to free memory associated with removed series sooner. (@rfratto)
-
Added a
disable_high_cardinality_metrics
configuration flag tootelcol
exporters and receivers to switch high cardinality debug metrics off. (@glindstedt) -
loki.source.kafka
component now exposes internal label__meta_kafka_offset
to indicate offset of consumed message. (@hainenber) -
Add a
tail_from_end
attribute inloki.source.file
to have the option to start tailing a file from the end if a cached position is not found. This is valuable when you want to tail a large file without reading its entire content. (@wildum) -
Flow: improve river config validation step in
prometheus.scrape
by comparingscrape_timeout
withscrape_interval
. (@wildum) -
Flow: add
randomization_factor
andmultiplier
to retry settings inotelcol
components. (@rfratto) -
Add support for
windows_certificate_filter
under http tls config block. (@mattdurham) -
Add
openstack
config converter to convert OpenStack yaml config (static mode) to river config (flow mode). (@wildum) -
Some
otelcol
components will now display their debug metrics via the Agent's/metrics
endpoint. Those components includeotelcol.receiver.otlp
,otelcol.exporter.otlp
andotelcol.processor.batch
. There may also be metrics from other components which are not documented yet. (@ptodev) -
Agent Management: Honor 503 ServiceUnavailable
Retry-After
header. (@jcreixell) -
Bump opentelemetry-collector and opentelemetry-collector-contrib versions from v0.80 to v0.85 (@wildum):
- add
authoriy
attribute tootelcol.exporter.loadbalancing
to override the default value in gRPC requests. - add
exemplars
support tootelcol.connector.spanmetrics
. - add
exclude_dimensions
attribute tootelcol.connector.spanmetrics
to exclude dimensions from the default set. - add
authority
attribute tootelcol.receiver.otlp
to override the default value in gRPC requests. - add
disable_keep_alives
attribute tootelcol.receiver.otlp
to disable the HTTP keep alive feature. - add
traces_url_path
,metrics_url_path
andlogs_url_path
attributes tootelcol.receiver.otlp
to specify the URl path to respectively receive traces, metrics and logs on. - add the value
json
to theencoding
attribute ofotelcol.receiver.kafka
. The component is now able to decodejson
payload and to insert it into the body of a log record.
- add
-
Added
scrape
block to customize the default behavior ofprometheus.operator.podmonitors
,prometheus.operator.probes
, andprometheus.operator.servicemonitors
. (@sberz) -
The
instance
label of targets exposed byprometheus.exporter.*
components is now more representative of what is being monitored. (@tpaschalis) -
Promtail converter will now treat
global positions configuration is not supported
as a Warning instead of Error. (@erikbaranowski) -
Add new
agent_component_dependencies_wait_seconds
histogram metric and a dashboard panel that measures how long components wait to be evaluated after their dependency is updated (@thampiotr) -
Add additional endpoint to debug scrape configs generated inside
prometheus.operator.*
components (@captncraig) -
Components evaluation is now performed in parallel, reducing the impact of slow components potentially blocking the entire telemetry pipeline. The
agent_component_evaluation_seconds
metric now measures evaluation time of each node separately, instead of all the directly and indirectly dependant nodes. (@thampiotr) -
Update Prometheus dependency to v2.46.0. (@tpaschalis)
-
The
client_secret
config argument in theotelcol.auth.oauth2
component is now of typesecret
instead of typestring
. (@ptodev)
-
Fixed
otelcol.exporter.prometheus
label names for theotel_scope_info
metric to match the OTLP Instrumentation Scope spec.name
is nowotel_scope_name
andversion
is nowotel_version_name
. (@erikbaranowski) -
Fixed a bug where converting
YACE
cloudwatch config to river skipped converting static jobs. (@berler) -
Fixed the
agent_prometheus_scrape_targets_gauge
incorrectly reporting all discovered targets instead of targets that belong to current instance when clustering is enabled. (@thampiotr) -
Fixed race condition in cleaning up metrics when stopping to tail files in static mode. (@thampiotr)
-
Fixed a bug where the BackOffLimit for the kubernetes tailer was always set to zero. (@anderssonw)
-
Fixed a bug where Flow agent fails to load
comment
statement inargument
block. (@hainenber) -
Fix initialization of the RAPL collector for the node_exporter integration and the prometheus.exporter.unix component. (@marctc)
-
Set instrumentation scope attribute for traces emitted by Flow component. (@hainenber)
-
Use Go 1.21.1 for builds. (@rfratto)
-
Read contextual attributes from Faro measurements (@codecapitano)
-
Rename Grafana Agent service in windows app and features to not include the description
-
Correct YAML level for
multitenancy_enabled
option in Mimir's config in examples. (@hainenber) -
Operator: Update default config reloader version. (@captncraig)
-
Sorting of common fields in log messages emitted by the agent in Flow mode have been standardized. The first fields will always be
ts
,level
, andmsg
, followed by non-common fields. Previously, the position ofmsg
was not consistent. (@rfratto) -
Documentation updated to link discovery.http and prometheus.scrape advanced configs (@proffalken)
-
Bump SNMP exporter version to v0.24.1 (@marctc)
-
Switch to
IBM/sarama
module. (@hainenber) -
Bump
webdevops/go-commons
to version containingLICENSE
. (@hainenber) -
prometheus.operator.probes
no longer ignores relabelingrule
blocks. (@sberz) -
Documentation updated to correct default path from
prometheus.exporter.windows
text_file
block (@timo1707) -
Bump
redis_exporter
to v1.54.0 (@spartan0x117) -
Migrate NodeJS installation in CI build image away from installation script. (@hainenber)
-
Fixed a bug where
otelcol.processor.discovery
could modify thetargets
passed by an upstream component. (@ptodev) -
Fixed a bug where
otelcol
components with a retry mechanism would not wait after the first retry. (@rfratto) -
Fixed a bug where documented default settings in
otelcol.exporter.loadbalancing
were never set. (@rfratto) -
Fix
loki.source.file
race condition in cleaning up metrics when stopping to tail files. (@thampiotr)
-
Restart managed components of a module loader only on if module content changes or the last load failed. This was specifically impacting
module.git
each time it pulls. (@erikbaranowski) -
Allow overriding default
User-Agent
forhttp.remote
component (@hainenber) -
Fix panic when running
grafana-agentctl config-check
against config files havingintegrations
block (both V1 and V2). (@hainenber) -
Fix a deadlock candidate in the
loki.process
component. (@tpaschalis) -
Fix an issue in the
eventhandler
integration where events would be double-logged: once by sending the event to Loki, and once by including the event in the Grafana Agent logs. Now, events are only ever sent to Loki. (@rfratto) -
Converters will now sanitize labels to valid River identifiers. (@erikbaranowski)
-
Converters will now return an Error diagnostic for unsupported
scrape_classic_histograms
andnative_histogram_bucket_limit
configs. (@erikbaranowski) -
Fix an issue in converters where targets of
discovery.relabel
components were repeating the first target for each source target instead of the correct target. (@erikbaranowski)
- Operator: Update default config reloader version. (@captncraig)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
-
loki.source.file
component will no longer automatically detect and decompress logs from compressed files. A new configuration block is available to enable decompression explicitly. See the upgrade guide for migration instructions. (@thampiotr) -
otelcol.exporter.prometheus
: Setinclude_scope_info
tofalse
by default. You can set it totrue
to preserve previous behavior. (@gouthamve) -
Set
retry_on_http_429
totrue
by default in thequeue_config
block in flow mode'sprometheus.remote_write
. (@wildum)
-
Add godeltaprof profiling types (
godeltaprof_memory
,godeltaprof_mutex
,godeltaprof_block
) topyroscope.scrape
component -
Flow: Allow the
logging
configuration block to tee the Agent's logs to one or more loki.* components. (@tpaschalis) -
Added support for
promtail
configuration conversion ingrafana-agent convert
andgrafana-agent run
commands. (@thampiotr) -
Flow: Add a new stage
non_indexed_labels
to attach non-indexed labels from extracted data to log line entry. (@vlad-diachenko) -
loki.write
now exposes basic WAL support. (@thepalbi) -
Flow: Users can now define
additional_fields
inloki.source.cloudflare
(@wildum) -
Flow: Added exemplar support for the
otelcol.exporter.prometheus
. (@wildum) -
Add a
labels
argument inloki.source.windowsevent
to associate additional labels with incoming logs. (@wildum) -
New Grafana Agent Flow components:
prometheus.exporter.gcp
- scrape GCP metrics. (@tburgessdev)otelcol.processor.span
- accepts traces telemetry data from otherotelcol
components and modifies the names and attributes of the spans. (@ptodev)discovery.uyuni
discovers scrape targets from a Uyuni Server. (@sparta0x117)discovery.eureka
discovers targets from a Eureka Service Registry. (@spartan0x117)discovery.openstack
- service discovery for OpenStack. (@marctc)discovery.hetzner
- service discovery for Hetzner Cloud. (@marctc)discovery.nomad
- service discovery from Nomad. (@captncraig)discovery.puppetdb
- service discovery from PuppetDB. (@captncraig)otelcol.processor.discovery
adds resource attributes to spans, where the attributes keys and values are sourced fromdiscovery.*
components. (@ptodev)otelcol.connector.spanmetrics
- creates OpenTelemetry metrics from traces. (@ptodev)
-
Integrations: include
direct_connect
,discovering_mode
andtls_basic_auth_config_path
fields for MongoDB configuration. (@gaantunes) -
Better validation of config file with
grafana-agentctl config-check
cmd (@fgouteroux) -
Integrations: make
udev
data path configurable in thenode_exporter
integration. (@sduranc) -
Clustering: Enable peer discovery with the go-discover package. (@tpaschalis)
-
Add
log_format
configuration to eventhandler integration and theloki.source.kubernetes_events
Flow component. (@sadovnikov) -
Allow
loki.source.file
to define the encoding of files. (@tpaschalis) -
Allow specification of
dimension_name_requirements
for Cloudwatch discovery exports. (@cvdv-au) -
Clustering: Enable nodes to periodically rediscover and rejoin peers. (@tpaschalis)
-
loki.write
WAL now exposes a last segment reclaimed metric. (@thepalbi) -
Update
memcached_exporter
tov0.13.0
, which includes bugfixes, new metrics, and the option to connect with TLS. (@spartan0x117) -
loki.write
now supports configuring retries on HTTP status code 429. (@wildum) -
Update
YACE
tov0.54.0
, which includes bugfixes for FIPS support. (@ashrayjain) -
Support decoupled scraping in the cloudwatch_exporter integration (@dtrejod).
-
Agent Management: Enable proxying support (@spartan0x117)
-
Update to config converter so default relabel
source_labels
are left off the river output. (@erikbaranowski) -
Rename
GrafanaAgentManagement
mixin rules toGrafanaAgentConfig
and update individual alerts to be more accurate. (@spartan0x117) -
Fix potential goroutine leak in log file tailing in static mode. (@thampiotr)
-
Fix issue on Windows where DNS short names were unresolvable. (@rfratto)
-
Fix panic in
prometheus.operator.*
when no Port supplied in Monitor crds. (@captncraig) -
Fix issue where Agent crashes when a blackbox modules config file is specified for blackbox integration. (@marctc)
-
Fix issue where the code from agent would not return to the Windows Service Manager (@jkroepke)
-
Fix issue where getting the support bundle failed due to using an HTTP Client that was not able to access the agent in-memory address. (@spartan0x117)
-
Fix an issue that lead the
loki.source.docker
container to use excessive CPU and memory. (@tpaschalis) -
Fix issue where
otelcol.exporter.loki
was not normalizing label names to comply with Prometheus conventions. (@ptodev) -
Agent Management: Fix issue where an integration defined multiple times could lead to undefined behaviour. (@jcreixell)
-
Sign RPMs with SHA256 for FIPs compatbility. (@mattdurham)
-
Fix issue where corrupt WAL segments lead to crash looping. (@tpaschalis)
-
Clarify usage documentation surrounding
loki.source.file
(@joshuapare)
-
Fix a bug which prevented the
app_agent_receiver
integration from processing traces. (@ptodev) -
(Agent static mode) Jaeger remote sampling works again, through a new
jaeger_remote_sampling
entry in the traces config. It is no longer configurable through the jaeger receiver. Support Jaeger remote sampling was removed accidentally in v0.35, and it is now restored, albeit via a different config entry. -
Clustering: Nodes take part in distributing load only after loading their component graph. (@tpaschalis)
-
Fix graceful termination when receiving SIGTERM/CTRL_SHUTDOWN_EVENT signals. (@tpaschalis)
-
Fix issue where the flow mode UI would show an empty page when navigating to an unhealthy
prometheus.operator
component or a healthyprometheus.operator
component which discovered no custom resources. (@rfratto) -
Fix panic when using
oauth2
without specifyingtls_config
. (@mattdurham) -
Fix issue where series records would never get written to the WAL if a scrape was rolled back, resulting in "dropped sample for series that was not explicitly dropped via relabelling" log messages. (@rfratto)
-
Fix RPM file digests so that installation on FIPS-enabled systems succeeds. (@andrewimeson)
- Compile journald support into builds of
grafana-agentctl
sografana-agentctl test-logs
functions as expected when testing tailing the systemd journal. (@rfratto)
- Fix incorrect display of trace IDs in the automatic_logging processor of static mode's traces subsystem. Users of the static mode's service graph processor are also advised to upgrade, although the bug should theoretically not affect them. (@ptodev)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
-
The algorithm for the "hash" action of
otelcol.processor.attributes
has changed. The change was made in PR #22831 of opentelemetry-collector-contrib. (@ptodev) -
otelcol.exporter.loki
now includes the instrumentation scope in its output. (@ptodev) -
otelcol.extension.jaeger_remote_sampling
removes the/
HTTP endpoint. The/sampling
endpoint is still functional. The change was made in PR #18070 of opentelemetry-collector-contrib. (@ptodev) -
The field
version
andauth
struct block fromwalk_params
inprometheus.exporter.snmp
and SNMP integration have been removed. The auth block now can be configured at top level, together withmodules
(@marctc) -
Rename
discovery.file
tolocal.file_match
to make it more clear that it discovers file on the local filesystem, and so it doesn't get confused with Prometheus' file discovery. (@rfratto) -
Remove the
discovery_target_decode
function in favor of using discovery components to better match the behavior of Prometheus' service discovery. (@rfratto) -
In the traces subsystem for Static mode, some metrics are removed and others are renamed. (@ptodev)
-
Removed metrics:
- "blackbox_exporter_config_last_reload_success_timestamp_seconds" (gauge)
- "blackbox_exporter_config_last_reload_successful" (gauge)
- "blackbox_module_unknown_total" (counter)
- "traces_processor_tail_sampling_count_traces_sampled" (counter)
- "traces_processor_tail_sampling_new_trace_id_received" (counter)
- "traces_processor_tail_sampling_sampling_decision_latency" (histogram)
- "traces_processor_tail_sampling_sampling_decision_timer_latency" (histogram)
- "traces_processor_tail_sampling_sampling_policy_evaluation_error" (counter)
- "traces_processor_tail_sampling_sampling_trace_dropped_too_early" (counter)
- "traces_processor_tail_sampling_sampling_traces_on_memory" (gauge)
- "traces_receiver_accepted_spans" (counter)
- "traces_receiver_refused_spans" (counter)
- "traces_exporter_enqueue_failed_log_records" (counter)
- "traces_exporter_enqueue_failed_metric_points" (counter)
- "traces_exporter_enqueue_failed_spans" (counter)
- "traces_exporter_queue_capacity" (gauge)
- "traces_exporter_queue_size" (gauge)
-
Renamed metrics:
- "traces_receiver_refused_spans" is renamed to "traces_receiver_refused_spans_total"
- "traces_receiver_accepted_spans" is renamed to "traces_receiver_refused_spans_total"
- "traces_exporter_sent_metric_points" is renamed to "traces_exporter_sent_metric_points_total"
-
-
The
remote_sampling
block has been removed fromotelcol.receiver.jaeger
. (@ptodev) -
(Agent static mode) Jaeger remote sampling used to be configured using the Jaeger receiver configuration. This receiver was updated to a new version, where support for remote sampling in the receiver was removed. Jaeger remote sampling is available as a separate configuration field starting in v0.35.3. (@ptodev)
otelcol.exporter.jaeger
has been deprecated and will be removed in Agent v0.38.0. (@ptodev)
-
The Pyroscope scrape component computes and sends delta profiles automatically when required to reduce bandwidth usage. (@cyriltovena)
-
Support
stage.geoip
inloki.process
. (@akselleirv) -
Integrations: Introduce the
squid
integration. (@armstrmi) -
Support custom fields in MMDB file for
stage.geoip
. (@akselleirv) -
Added json_path function to river stdlib. (@jkroepke)
-
Add
format
,join
,tp_lower
,replace
,split
,trim
,trim_prefix
,trim_suffix
,trim_space
,to_upper
functions to river stdlib. (@jkroepke) -
Flow UI: Add a view for listing the Agent's peers status when clustering is enabled. (@tpaschalis)
-
Add a new CLI command
grafana-agent convert
for converting a river file from supported formats to river. (@erikbaranowski) -
Add support to the
grafana-agent run
CLI for converting a river file from supported formats to river. (@erikbaranowski) -
Add boringcrypto builds and docker images for Linux arm64 and x64. (@mattdurham)
-
New Grafana Agent Flow components:
discovery.file
discovers scrape targets from files. (@spartan0x117)discovery.kubelet
collect scrape targets from the Kubelet API. (@gcampbell12)module.http
runs a Grafana Agent Flow module loaded from a remote HTTP endpoint. (@spartan0x117)otelcol.processor.attributes
accepts telemetry data from otherotelcol
components and modifies attributes of a span, log, or metric. (@ptodev)prometheus.exporter.cloudwatch
- scrape AWS CloudWatch metrics (@thepalbi)prometheus.exporter.elasticsearch
collects metrics from Elasticsearch. (@marctc)prometheus.exporter.kafka
collects metrics from Kafka Server. (@oliver-zhang)prometheus.exporter.mongodb
collects metrics from MongoDB. (@marctc)prometheus.exporter.squid
collects metrics from a squid server. (@armstrmi)prometheus.operator.probes
- discovers Probe resources in your Kubernetes cluster and scrape the targets they reference. (@captncraig)pyroscope.ebpf
collects system-wide performance profiles from the current host (@korniltsev)otelcol.exporter.loadbalancing
- export traces and logs to multiple OTLP gRPC endpoints in a load-balanced way. (@ptodev)
-
New Grafana Agent Flow command line utilities:
grafana-agent tools prometheus.remote_write
holds a collection of remote write-specific tools. These have been ported over from theagentctl
command. (@rfratto)
-
A new
action
argument forotelcol.auth.headers
. (@ptodev) -
New
metadata_keys
andmetadata_cardinality_limit
arguments forotelcol.processor.batch
. (@ptodev) -
New
boolean_attribute
andottl_condition
sampling policies forotelcol.processor.tail_sampling
. (@ptodev) -
A new
initial_offset
argument forotelcol.receiver.kafka
. (@ptodev)
-
Attributes and blocks set to their default values will no longer be shown in the Flow UI. (@rfratto)
-
Tanka config: retain cAdvisor metrics for system processes (Kubelet, Containerd, etc.) (@bboreham)
-
Update cAdvisor dependency to v0.47.0. (@jcreixell)
-
Upgrade and improve Cloudwatch exporter integration (@thepalbi)
-
Update
node_exporter
dependency to v1.6.0. (@spartan0x117) -
Enable
prometheus.relabel
to work with Prometheus' Native Histograms. (@tpaschalis) -
Update
dnsmasq_exporter
to last version. (@marctc) -
Add deployment spec options to describe operator's Prometheus Config Reloader image. (@alekseybb197)
-
Update
module.git
with basic and SSH key authentication support. (@djcode) -
Support
clustering
block inprometheus.operator.servicemonitors
andprometheus.operator.podmonitors
components to distribute targets amongst clustered agents. (@captncraig) -
Update
redis_exporter
dependency to v1.51.0. (@jcreixell) -
The Grafana Agent mixin now includes a dashboard for the logs pipeline. (@thampiotr)
-
The Agent Operational dashboard of Grafana Agent mixin now has more descriptive panel titles, Y-axis units
-
Add
write_relabel_config
toprometheus.remote_write
(@jkroepke) -
Update OpenTelemetry Collector dependencies from v0.63.0 to v0.80.0. (@ptodev)
-
Allow setting the node name for clustering with a command-line flag. (@tpaschalis)
-
Allow
prometheus.exporter.snmp
and SNMP integration to be configured passing a YAML block. (@marctc) -
Some metrics have been added to the traces subsystem for Static mode. (@ptodev)
- "traces_processor_batch_batch_send_size" (histogram)
- "traces_processor_batch_batch_size_trigger_send_total" (counter)
- "traces_processor_batch_metadata_cardinality" (gauge)
- "traces_processor_batch_timeout_trigger_send_total" (counter)
- "traces_rpc_server_duration" (histogram)
- "traces_exporter_send_failed_metric_points_total" (counter)
- "traces_exporter_send_failed_spans_total" (counter)
- "traces_exporter_sent_spans_total" (counter)
-
Added support for custom
length
time setting in Cloudwatch component and integration. (@thepalbi)
-
Fix issue where
remote.http
incorrectly had a status of "Unknown" until the period specified by the polling frquency elapsed. (@rfratto) -
Add signing region to remote.s3 component for use with custom endpoints so that Authorization Headers work correctly when proxying requests. (@mattdurham)
-
Fix oauth default scope in
loki.source.azure_event_hubs
. (@akselleirv) -
Fix bug where
otelcol.exporter.otlphttp
ignores configuration fortraces_endpoint
,metrics_endpoint
, andlogs_endpoint
attributes. (@SimoneFalzone) -
Fix issue in
prometheus.remote_write
where thequeue_config
andmetadata_config
blocks used incorrect defaults when not specified in the config file. (@rfratto) -
Fix issue where published RPMs were not signed. (@rfratto)
-
Fix issue where flow mode exports labeled as "string or secret" could not be used in a binary operation. (@rfratto)
-
Fix Grafana Agent mixin's "Agent Operational" dashboard expecting pods to always have
grafana-agent-.*
prefix. (@thampiotr) -
Change the HTTP Path and Data Path from the controller-local ID to the global ID for components loaded from within a module loader. (@spartan0x117)
-
Fix bug where
stage.timestamp
inloki.process
wasn't able to correctly parse timezones. This issue only impacts the dedicatedgrafana-agent-flow
binary. (@rfratto) -
Fix bug where JSON requests to
loki.source.api
would not be handled correctly. This adds/loki/api/v1/raw
and/loki/api/v1/push
endpoints toloki.source.api
and maps the/api/v1/push
and/api/v1/raw
to the/loki
prefixed endpoints. (@mattdurham) -
Upgrade
loki.write
dependencies to latest changes. (@thepalbi)
-
Mongodb integration has been re-enabled. (@jcreixell, @marctc)
-
Build with go 1.20.6 (@captncraig)
-
Clustering for Grafana Agent in flow mode has graduated from experimental to beta.
- Fixes a bug in conversion of OpenTelemetry histograms when exported to Prometheus. (@grcevski)
- Enforce sha256 digest signing for rpms enabling installation on FIPS-enabled OSes. (@kfriedrich123)
- Fix panic from improper startup ordering in
prometheus.operator.servicemonitors
. (@captncraig)
- Replace map cache in prometheus.relabel with an LRU cache. (@mattdurham)
- Integrations: Extend
statsd
integration to configure relay endpoint. (@arminaaki)
- Fix a bug where
prometheus.relabel
would not correctly relabel when there is a cache miss. (@thampiotr) - Fix a bug where
prometheus.relabel
would not correctly relabel exemplars or metadata. (@tpaschalis) - Fixes several issues with statsd exporter. (@jcreixell, @marctc)
- Mongodb integration has been disabled for the time being due to licensing issues. (@jcreixell)
-
Fixed application of sub-collector defaults using the
windows_exporter
integration orprometheus.exporter.windows
. (@mattdurham) -
Fix issue where
remote.http
did not fail early if the initial request failed. This caused failed requests to initially export empty values, which could lead to propagating issues downstream to other components which expect the export to be non-empty. (@rfratto) -
Allow
bearerTokenFile
field to be used in ServiceMonitors. (@captncraig) -
Fix issue where metrics and traces were not recorded from components within modules. (@mattdurham)
-
service_name
label is inferred from discovery meta labels inpyroscope.scrape
(@korniltsev)
- Add logging to failed requests in
remote.http
. (@rfratto)
-
The experimental dynamic configuration feature has been removed in favor of Flow mode. (@mattdurham)
-
The
oracledb
integration configuration has removed a redundant fieldmetrics_scrape_interval
. Use thescrape_interval
parameter of the integration if a custom scrape interval is required. (@schmikei) -
Upgrade the embedded windows_exporter to commit 79781c6. (@jkroepke)
-
Prometheus exporters in Flow mode now set the
instance
label to a value similar to the one they used to have in Static mode ( by default, customized by some integrations). (@jcreixell) -
phlare.scrape
andphlare.write
have been renamed topyroscope.scrape
andpyroscope.scrape
. (@korniltsev)
-
New Grafana Agent Flow components:
loki.source.api
- receive Loki log entries over HTTP (e.g. from other agents). (@thampiotr)prometheus.operator.servicemonitors
discovers ServiceMonitor resources in your Kubernetes cluster and scrape the targets they reference. (@captncraig, @marctc, @jcreixell)prometheus.receive_http
- receive Prometheus metrics over HTTP (e.g. from other agents). (@thampiotr)remote.vault
retrieves a secret from Vault. (@rfratto)prometheus.exporter.snowflake
collects metrics from a snowflake database (@jonathanWamsley)prometheus.exporter.mssql
collects metrics from Microsoft SQL Server (@jonathanwamsley)prometheus.exporter.oracledb
collects metrics from oracledb (@jonathanwamsley)prometheus.exporter.dnsmasq
collects metrics from a dnsmasq server. (@spartan0x117)loki.source.awsfirehose
- receive Loki log entries from AWS Firehose via HTTP (@thepalbi)discovery.http
service discovery via http. (@captncraig)
-
Added new functions to the River standard library:
coalesce
returns the first non-zero value from a list of arguments. (@jkroepke)nonsensitive
converts a River secret back into a string. (@rfratto)
-
Support to attach node metadata to pods and endpoints targets in
discovery.kubernetes
. (@laurovenancio) -
Support ability to add optional custom headers to
loki.write
endpoint block (@aos) -
Support in-memory HTTP traffic for Flow components.
prometheus.exporter
components will now export a target containing an internal HTTP address.prometheus.scrape
, when given that internal HTTP address, will connect to the server in-memory, bypassing the network stack. Use the new--server.http.memory-addr
flag to customize which address is used for in-memory traffic. (@rfratto) -
Disable node_exporter on Windows systems (@jkroepke)
-
Operator support for OAuth 2.0 Client in LogsClientSpec (@DavidSpek)
-
Support
clustering
block inphlare.scrape
components to distribute targets amongst clustered agents. (@rfratto) -
Delete stale series after a single WAL truncate instead of two. (@rfratto)
-
Update OracleDB Exporter dependency to 0.5.0 (@schmikei)
-
Embed Google Fonts on Flow UI (@jkroepke)
-
Enable Content-Security-Policies on Flow UI (@jkroepke)
-
Update azure-metrics-exporter to v0.0.0-20230502203721-b2bfd97b5313 (@kgeckhart)
-
Update azidentity dependency to v1.3.0. (@akselleirv)
-
Add custom labels to journal entries in
loki.source.journal
(@sbhrule15) -
prometheus.operator.podmonitors
andprometheus.operator.servicemonitors
can now access cluster secrets for authentication to targets. (@captncraig)
-
Fix
loki.source.(gcplog|heroku)
http
andgrpc
blocks were overriding defaults with zero-values on non-present fields. (@thepalbi) -
Fix an issue where defining
logging
ortracing
blocks inside of a module would generate a panic instead of returning an error. (@erikbaranowski) -
Fix an issue where not specifying either
http
norgrpc
blocks could result in a panic forloki.source.heroku
andloki.source.gcplog
components. (@thampiotr) -
Fix an issue where build artifacts for IBM S390x were being built with the GOARCH value for the PPC64 instead. (tpaschalis)
-
Fix an issue where the Grafana Agent Flow RPM used the wrong path for the environment file, preventing the service from loading. (@rfratto)
-
Fix an issue where the cluster advertise address was overwriting the join addresses. (@laurovenancio)
-
Fix targets deduplication when clustering mode is enabled. (@laurovenancio)
-
Fix issue in operator where any version update will restart all agent pods simultaneously. (@captncraig)
-
Fix an issue where
loki.source.journald
did not create the positions directory with the appropriate permissions. (@tpaschalis) -
Fix an issue where fanning out log entries to multiple
loki.process
components lead to a race condition. (@tpaschalis) -
Fix panic in
prometheus.operator.servicemonitors
from relabel rules without certain defaults. (@captncraig) -
Fix issue in modules export cache throwing uncomparable errors. (@mattdurham)
-
Fix issue where the UI could not navigate to components loaded by modules. (@rfratto)
-
Fix issue where using exporters inside modules failed due to not passing the in-memory address dialer. (@mattdurham)
-
Add signing region to remote.s3 component for use with custom endpoints so that Authorization Headers work correctly when proxying requests. (@mattdurham)
-
Fix missing
instance
key forprometheus.exporter.dnsmasq
component. (@spartan0x117)
-
Add metrics when clustering mode is enabled. (@rfratto)
-
Document debug metric
loki_process_dropped_lines_by_label_total
in loki.process. (@akselleirv) -
Add
agent_wal_out_of_order_samples_total
metric to track samples received out of order. (@rfratto) -
Add CLI flag
--server.http.enable-pprof
to grafana-agent-flow to conditionally enable/debug/pprof
endpoints (@jkroepke) -
Use Go 1.20.4 for builds. (@tpaschalis)
-
Integrate the new ExceptionContext which was recently added to the Faro Web-SDK in the app_agent_receiver Payload. (@codecapitano)
-
Flow clustering: clusters will now use 512 tokens per node for distributing work, leading to better distribution. However, rolling out this change will cause some incorrerct or missing assignments until all nodes are updated. (@rfratto)
-
Change the Docker base image for Linux containers to
ubuntu:lunar
. (@rfratto)
-
Fix issue where component evaluation time was overridden by a "default health" message. (@rfratto)
-
Honor timeout when trying to establish a connection to another agent in Flow clustering mode. (@rfratto)
-
Fix an issue with the grafana/agent windows docker image entrypoint not targeting the right location for the config. (@erikbaranowski)
-
Fix issue where the
node_exporter
integration andprometheus.exporter.unix
diskstat_device_include
component could not set the allowlist field for the diskstat collector. (@tpaschalis) -
Fix an issue in
loki.source.heroku
where updating thelabels
oruse_incoming_timestamp
would not take effect. (@thampiotr) -
Flow: Fix an issue within S3 Module where the S3 path was not parsed correctly when the path consists of a parent directory. (@jastisriradheshyam)
-
Flow: Fix an issue on Windows where
prometheus.remote_write
failed to read WAL checkpoints. This issue led to memory leaks once the initial checkpoint was created, and prevented a fresh process from being able to deliver metrics at all. (@rfratto) -
Fix an issue where the
loki.source.kubernetes
component could lead to the Agent crashing due to a race condition. (@tpaschalis)
- The
phlare.scrape
Flow componentfetch profile failed
log has been set todebug
instead oferror
. (@erikbaranowski)
-
Fix spelling of the
frequency
argument on thelocal.file
component. (@tpaschalis) -
Fix bug where some capsule values (such as Prometheus receivers) could not properly be used as an argument to a module. (@rfratto)
-
Fix version information not displaying correctly when passing the
--version
flag or in theagent_build_info
metric. (@rfratto) -
Fix issue in
loki.source.heroku
andloki.source.gcplog
where updating the component would cause Grafana Agent Flow's Prometheus metrics endpoint to return an error until the process is restarted. (@rfratto) -
Fix issue in
loki.source.file
where updating the component caused goroutines to leak. (@rfratto)
- Support Bundles report the status of discovered log targets. (@tpaschalis)
-
Support for 32-bit ARM builds is removed for the foreseeable future due to Go compiler issues. We will consider bringing back 32-bit ARM support once our Go compiler issues are resolved and 32-bit ARM builds are stable. (@rfratto)
-
Agent Management:
agent_management.api_url
config field has been replaced byagent_management.host
. The API path and version is now defined by the Agent. (@jcreixell) -
Agent Management:
agent_management.protocol
config field now allows defining "http" and "https" explicitly. Previously, "http" was previously used for both, with the actual protocol used inferred from the api url, which led to confusion. When upgrading, make sure to set to "https" when replacingapi_url
withhost
. (@jcreixell) -
Agent Management:
agent_management.remote_config_cache_location
config field has been replaced byagent_management.remote_configuration.cache_location
. (@jcreixell) -
Remove deprecated symbolic links to to
/bin/agent*
in Docker containers, as planned in v0.31. (@tpaschalis)
- Dynamic Configuration will be removed in v0.34. Grafana Agent Flow supersedes this functionality. (@mattdurham)
-
New Grafana Agent Flow components:
discovery.dns
DNS service discovery. (@captncraig)discovery.ec2
service discovery for aws ec2. (@captncraig)discovery.lightsail
service discovery for aws lightsail. (@captncraig)discovery.gce
discovers resources on Google Compute Engine (GCE). (@marctc)discovery.digitalocean
provides service discovery for DigitalOcean. (@spartan0x117)discovery.consul
service discovery for Consul. (@jcreixell)discovery.azure
provides service discovery for Azure. (@spartan0x117)module.file
runs a Grafana Agent Flow module loaded from a file on disk. (@erikbaranowski)module.git
runs a Grafana Agent Flow module loaded from a file within a Git repository. (@rfratto)module.string
runs a Grafana Agent Flow module passed to the component by an expression containing a string. (@erikbaranowski, @rfratto)otelcol.auth.oauth2
performs OAuth 2.0 authentication for HTTP and gRPC based OpenTelemetry exporters. (@ptodev)otelcol.extension.jaeger_remote_sampling
provides an endpoint from which to pull Jaeger remote sampling documents. (@joe-elliott)otelcol.exporter.logging
accepts OpenTelemetry data from otherotelcol
components and writes it to the console. (@erikbaranowski)otelcol.auth.sigv4
performs AWS Signature Version 4 (SigV4) authentication for making requests to AWS services viaotelcol
components that support authentication extensions. (@ptodev)prometheus.exporter.blackbox
collects metrics from Blackbox exporter. (@marctc)prometheus.exporter.mysql
collects metrics from a MySQL database. (@spartan0x117)prometheus.exporter.postgres
collects metrics from a PostgreSQL database. (@spartan0x117)prometheus.exporter.statsd
collects metrics from a Statsd instance. (@gaantunes)prometheus.exporter.snmp
collects metrics from SNMP exporter. (@marctc)prometheus.operator.podmonitors
discovers PodMonitor resources in your Kubernetes cluster and scrape the targets they reference. (@captncraig, @marctc, @jcreixell)prometheus.exporter.windows
collects metrics from a Windows instance. (@jkroepke)prometheus.exporter.memcached
collects metrics from a Memcached server. (@spartan0x117)loki.source.azure_event_hubs
reads messages from Azure Event Hub using Kafka and forwards them to otherloki
components. (@akselleirv)
-
Add support for Flow-specific system packages:
- Flow-specific DEB packages. (@rfratto, @robigan)
- Flow-specific RPM packages. (@rfratto, @robigan)
- Flow-specific macOS Homebrew Formula. (@rfratto)
- Flow-specific Windows installer. (@rfratto)
The Flow-specific packages allow users to install and run Grafana Agent Flow alongside an existing installation of Grafana Agent.
-
Agent Management: Add support for integration snippets. (@jcreixell)
-
Flow: Introduce a gossip-over-HTTP/2 clustered mode.
prometheus.scrape
component instances can opt-in to distributing scrape load between cluster peers. (@tpaschalis)
-
Flow: Add retries with backoff logic to Phlare write component. (@cyriltovena)
-
Operator: Allow setting runtimeClassName on operator-created pods. (@captncraig)
-
Operator: Transparently compress agent configs to stay under size limitations. (@captncraig)
-
Update Redis Exporter Dependency to v1.49.0. (@spartan0x117)
-
Update Loki dependency to the k144 branch. (@andriikushch)
-
Flow: Add OAUTHBEARER mechanism to
loki.source.kafka
using Azure as provider. (@akselleirv) -
Update Process Exporter dependency to v0.7.10. (@spartan0x117)
-
Agent Management: Introduces backpressure mechanism for remote config fetching (obeys 429 request
Retry-After
header). (@spartan0x117) -
Flow: support client TLS settings (CA, client certificate, client key) being provided from other components for the following components:
discovery.docker
discovery.kubernetes
loki.source.kafka
loki.source.kubernetes
loki.source.podlogs
loki.write
mimir.rules.kubernetes
otelcol.auth.oauth2
otelcol.exporter.jaeger
otelcol.exporter.otlp
otelcol.exporter.otlphttp
otelcol.extension.jaeger_remote_sampling
otelcol.receiver.jaeger
otelcol.receiver.kafka
phlare.scrape
phlare.write
prometheus.remote_write
prometheus.scrape
remote.http
-
Flow: support server TLS settings (client CA, server certificate, server key) being provided from other components for the following components:
loki.source.syslog
otelcol.exporter.otlp
otelcol.extension.jaeger_remote_sampling
otelcol.receiver.jaeger
otelcol.receiver.opencensus
otelcol.receiver.zipkin
-
Flow: Define custom http method and headers in
remote.http
component (@jkroepke) -
Flow: Add config property to
prometheus.exporter.blackbox
to define the config inline (@jkroepke) -
Update Loki Dependency to k146 which includes configurable file watchers (@mattdurham)
-
Flow: fix issue where Flow would return an error when trying to access a key of a map whose value was the zero value (
null
,0
,false
,[]
,{}
). Whether an error was returned depended on the internal type of the value. (@rfratto) -
Flow: fix issue where using the
jaeger_remote
sampler for thetracing
block would fail to parse the response from the remote sampler server if it used strings for the strategy type. This caused sampling to fall back to the default rate. (@rfratto) -
Flow: fix issue where components with no arguments like
loki.echo
were not viewable in the UI. (@rfratto) -
Flow: fix deadlock in
loki.source.file
where terminating tailers would hang while flushing remaining logs, preventingloki.source.file
from being able to update. (@rfratto) -
Flow: fix deadlock in
loki.process
where a component with no stages would hang forever on handling logs. (@rfratto) -
Fix issue where a DefaultConfig might be mutated during unmarshaling. (@jcreixell)
-
Fix issues where CloudWatch Exporter cannot use FIPS Endpoints outside of USA regions (@aglees)
-
Fix issue where scraping native Prometheus histograms would leak memory. (@rfratto)
-
Flow: fix issue where
loki.source.docker
component could deadlock. (@tpaschalis) -
Flow: fix issue where
prometheus.remote_write
created unnecessary extra child directories to store the WAL in. (@rfratto) -
Fix internal metrics reported as invalid by promtool's linter. (@tpaschalis)
-
Fix issues with cri stage which treats partial line coming from any stream as same. (@kavirajk @aglees)
-
Operator: fix for running multiple operators with different
--agent-selector
flags. (@captncraig) -
Operator: respect FilterRunning on PodMonitor and ServiceMonitor resources to only scrape running pods. (@captncraig)
-
Fixes a bug where the github exporter would get stuck in an infinite loop under certain conditions. (@jcreixell)
-
Fix bug where
loki.source.docker
always failed to start. (@rfratto)
-
Grafana Agent Docker containers and release binaries are now published for s390x. (@rfratto)
-
Use Go 1.20.3 for builds. (@rfratto)
-
Change the Docker base image for Linux containers to
ubuntu:kinetic
. (@rfratto) -
Update prometheus.remote_write defaults to match new prometheus remote-write defaults. (@erikbaranowski)
-
Flow: Fixes slow reloading of targets in
phlare.scrape
component. (@cyriltovena) -
Flow: add a maximum connection lifetime of one hour when tailing logs from
loki.source.kubernetes
andloki.source.podlogs
to recover from an issue where the Kubernetes API server stops responding with logs without closing the TCP connection. (@rfratto) -
Flow: fix issue in
loki.source.kubernetes
where__pod__uid__
meta label defaulted incorrectly to the container name, causing tailers to never restart. (@rfratto)
-
Support for the embedded Flow UI for 32-bit ARMv6 builds is temporarily removed. (@rfratto)
-
Node Exporter configuration options changed to align with new upstream version (@Thor77):
diskstats_ignored_devices
is nowdiskstats_device_exclude
in agent configuration.ignored_devices
is nowdevice_exclude
in flow configuration.
-
Some blocks in Flow components have been merged with their parent block to make the block hierarchy smaller:
discovery.docker > http_client_config
is merged into thediscovery.docker
block. (@erikbaranowski)discovery.kubernetes > http_client_config
is merged into thediscovery.kubernetes
block. (@erikbaranowski)loki.source.kubernetes > client > http_client_config
is merged into theclient
block. (@erikbaranowski)loki.source.podlogs > client > http_client_config
is merged into theclient
block. (@erikbaranowski)loki.write > endpoint > http_client_config
is merged into theendpoint
block. (@erikbaranowski)mimir.rules.kubernetes > http_client_config
is merged into themimir.rules.kubernetes
block. (@erikbaranowski)otelcol.receiver.opencensus > grpc
is merged into theotelcol.receiver.opencensus
block. (@ptodev)otelcol.receiver.zipkin > http
is merged into theotelcol.receiver.zipkin
block. (@ptodev)phlare.scrape > http_client_config
is merged into thephlare.scrape
block. (@erikbaranowski)phlare.write > endpoint > http_client_config
is merged into theendpoint
block. (@erikbaranowski)prometheus.remote_write > endpoint > http_client_config
is merged into theendpoint
block. (@erikbaranowski)prometheus.scrape > http_client_config
is merged into theprometheus.scrape
block. (@erikbaranowski)
-
The
loki.process
component now uses a combined name for stages, simplifying the block hierarchy. For example, thestage > json
block hierarchy is now a single block calledstage.json
. All stage blocks inloki.process
have been updated to use this simplified hierarchy. (@tpaschalis) -
remote.s3
client_options
block has been renamed toclient
. (@mattdurham) -
Renamed
prometheus.integration.node_exporter
toprometheus.exporter.unix
. (@jcreixell) -
As first announced in v0.30, support for the
EXPERIMENTAL_ENABLE_FLOW
environment variable has been removed in favor ofAGENT_MODE=flow
. (@rfratto)
-
New integrations:
oracledb
(@schmikei)mssql
(@binaryfissiongames)cloudwatch metrics
(@thepalbi)azure
(@kgeckhart)gcp
(@kgeckhart, @ferruvich)
-
New Grafana Agent Flow components:
loki.echo
writes received logs to stdout. (@tpaschalis, @rfratto)loki.source.docker
reads logs from Docker containers and forwards them to otherloki
components. (@tpaschalis)loki.source.kafka
reads logs from Kafka events and forwards them to otherloki
components. (@erikbaranowski)loki.source.kubernetes_events
watches for Kubernetes Events and converts them into log lines to forward to otherloki
components. It is the equivalent of theeventhandler
integration. (@rfratto)otelcol.processor.tail_sampling
samples traces based on a set of defined policies fromotelcol
components before forwarding them to otherotelcol
components. (@erikbaranowski)prometheus.exporter.apache
collects metrics from an apache web server (@captncraig)prometheus.exporter.consul
collects metrics from a consul installation (@captncraig)prometheus.exporter.github
collects metrics from GitHub (@jcreixell)prometheus.exporter.process
aggregates and collects metrics by scraping/proc
. (@spartan0x117)prometheus.exporter.redis
collects metrics from a redis database (@spartan0x117)
-
Flow: Support
keepequal
anddropequal
actions for relabeling. (@cyriltovena) -
Update Prometheus Node Exporter integration to v1.5.0. (@Thor77)
-
Grafana Agent Flow will now reload the config file when
SIGHUP
is sent to the process. (@rfratto) -
If using the official RPM and DEB packages for Grafana Agent, invoking
systemctl reload grafana-agent
will now reload the configuration file. (@rfratto) -
Flow: the
loki.process
component now implements all the same processing stages as Promtail's pipelines. (@tpaschalis) -
Flow: new metric for
prometheus.scrape
-agent_prometheus_scrape_targets_gauge
. (@ptodev) -
Flow: new metric for
prometheus.scrape
andprometheus.relabel
-agent_prometheus_forwarded_samples_total
. (@ptodev) -
Flow: add
constants
into the standard library to expose the hostname, OS, and architecture of the system Grafana Agent is running on. (@rfratto) -
Flow: add timeout to loki.source.podlogs controller setup. (@polyrain)
-
Fixed a reconciliation error in Grafana Agent Operator when using
tlsConfig
onProbe
. (@supergillis) -
Fix issue where an empty
server:
config stanza would cause debug-level logging. An emptyserver:
is considered a misconfiguration, and thus will error out. (@neomantra) -
Flow: fix an error where some error messages that crossed multiple lines added extra an extra
|
character when displaying the source file on the starting line. (@rfratto) -
Flow: fix issues in
agent fmt
where adding an inline comment on the same line as a[
or{
would cause indentation issues on subsequent lines. (@rfratto) -
Flow: fix issues in
agent fmt
where line comments in arrays would be given the wrong identation level. (@rfratto) -
Flow: fix issues with
loki.file
andloki.process
where deadlock contention or logs fail to process. (@mattdurham) -
Flow:
oauth2 > tls_config
was documented as a block but coded incorrectly as an attribute. This is now a block in code. This impacteddiscovery.docker
,discovery.kubernetes
,loki.source.kubernetes
,loki.write
,mimir.rules.kubernetes
,phlare.scrape
,phlare.write
,prometheus.remote_write
,prometheus.scrape
, andremote.http
(@erikbaranowski) -
Flow: Fix issue where using
river:",label"
causes the UI to return nothing. (@mattdurham)
-
Use Go 1.20 for builds. (@rfratto)
-
The beta label from Grafana Agent Flow has been removed. A subset of Flow components are still marked as beta or experimental:
loki.echo
is explicitly marked as beta.loki.source.kubernetes
is explicitly marked as experimental.loki.source.podlogs
is explicitly marked as experimental.mimir.rules.kubernetes
is explicitly marked as beta.otelcol.processor.tail_sampling
is explicitly marked as beta.otelcol.receiver.loki
is explicitly marked as beta.otelcol.receiver.prometheus
is explicitly marked as beta.phlare.scrape
is explicitly marked as beta.phlare.write
is explicitly marked as beta.
-
loki.source.cloudflare
: fix issue where thezone_id
argument was being ignored, and theapi_token
argument was being used for the zone instead. (@rfratto) -
loki.source.cloudflare
: fix issue whereapi_token
argument was not marked as a sensitive field. (@rfratto)
- In the Agent Operator, upgrade the
prometheus-config-reloader
dependency from version 0.47.0 to version 0.62.0. (@ptodev)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
- All release Windows
.exe
files are now published as a zip archive. Previously,grafana-agent-installer.exe
was unzipped. (@rfratto)
- Support Go 1.20 for builds. Official release binaries are still produced using Go 1.19. (@rfratto)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
-
Release binaries (including inside Docker containers) have been renamed to be prefixed with
grafana-
(@rfratto):agent
is nowgrafana-agent
.agentctl
is nowgrafana-agentctl
.agent-operator
is nowgrafana-agent-operator
.
- A symbolic link in Docker containers from the old binary name to the new binary name has been added. These symbolic links will be removed in v0.33. (@rfratto)
-
New Grafana Agent Flow components:
loki.source.cloudflare
reads logs from Cloudflare's Logpull API and forwards them to otherloki
components. (@tpaschalis)loki.source.gcplog
reads logs from GCP cloud resources using Pub/Sub subscriptions and forwards them to otherloki
components. (@tpaschalis)loki.source.gelf
listens for Graylog logs. (@mattdurham)loki.source.heroku
listens for Heroku messages over TCP a connection and forwards them to otherloki
components. (@erikbaranowski)loki.source.journal
read messages from systemd journal. (@mattdurham)loki.source.kubernetes
collects logs from Kubernetes pods using the Kubernetes API. (@rfratto)loki.source.podlogs
discovers PodLogs resources on Kubernetes and uses the Kubernetes API to collect logs from the pods specified by the PodLogs resource. (@rfratto)loki.source.syslog
listens for Syslog messages over TCP and UDP connections and forwards them to otherloki
components. (@tpaschalis)loki.source.windowsevent
reads logs from Windows Event Log. (@mattdurham)otelcol.exporter.jaeger
forwards OpenTelemetry data to a Jaeger server. (@erikbaranowski)otelcol.exporter.loki
forwards OTLP-formatted data to compatibleloki
receivers. (@tpaschalis)otelcol.receiver.kafka
receives telemetry data from Kafka. (@rfratto)otelcol.receiver.loki
receives Loki logs, converts them to the OTLP log format and forwards them to otherotelcol
components. (@tpaschalis)otelcol.receiver.opencensus
receives OpenConsensus-formatted traces or metrics. (@ptodev)otelcol.receiver.zipkin
receives Zipkin-formatted traces. (@rfratto)phlare.scrape
collects application performance profiles. (@cyriltovena)phlare.write
sends application performance profiles to Grafana Phlare. (@cyriltovena)mimir.rules.kubernetes
discoversPrometheusRule
Kubernetes resources and loads them into a Mimir instance. (@Logiraptor)
-
Flow components which work with relabeling rules (
discovery.relabel
,prometheus.relabel
andloki.relabel
) now export a new value named Rules. This value returns a copy of the currently configured rules. (@tpaschalis) -
New experimental feature: agent-management. Polls configured remote API to fetch new configs. (@spartan0x117)
-
Introduce global configuration for logs. (@jcreixell)
-
Handle faro-web-sdk
View
meta in app_agent_receiver. (@rlankfo) -
Flow: the targets in debug info from
loki.source.file
are now individual blocks. (@rfratto) -
Grafana Agent Operator: add promtail limit stage to the operator. (@spartan0x117)
-
Flow UI: Fix the issue with messy layout on the component list page while browser window resize (@xiyu95)
-
Flow UI: Display the values of all attributes unless they are nil. (@ptodev)
-
Flow:
prometheus.relabel
andprometheus.remote_write
will now error if they have exited. (@ptodev) -
Flow: Fix issue where negative numbers would convert to floating-point values incorrectly, treating the sign flag as part of the number. (@rfratto)
-
Flow: fix a goroutine leak when
loki.source.file
is passed more than one target with identical set of public labels. (@rfratto) -
Fix issue where removing and re-adding log instance configurations causes an error due to double registration of metrics (@spartan0x117, @jcreixell)
-
Use Go 1.19.4 for builds. (@erikbaranowski)
-
New windows containers for agent and agentctl. These can be found moving forward with the ${Version}-windows tags for grafana/agent and grafana/agentctl docker images (@erikbaranowski)
-
Flow:
prometheus.relabel
will no longer modify the labels of the original metrics, which could lead to the incorrect application of relabel rules on subsequent relabels. (@rfratto) -
Flow:
loki.source.file
will no longer deadlock other components if log lines cannot be sent to Loki.loki.source.file
will wait for 5 seconds per file to finish flushing read logs to the client, after which it will drop them, resulting in lost logs. (@rfratto) -
Operator: Fix the handling of the enableHttp2 field as a boolean in
pod_monitor
andservice_monitor
templates. (@tpaschalis)
-
Fix issue where journald support was accidentally removed. (@tpaschalis)
-
Fix issue where some traces' metrics where not collected. (@marctc)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
- The
ebpf_exporter
integration has been removed due to issues with static linking. It may be brought back once these are resolved. (@tpaschalis)
- The
EXPERIMENTAL_ENABLE_FLOW
environment variable is deprecated in favor ofAGENT_MODE=flow
. Support forEXPERIMENTAL_ENABLE_FLOW
will be removed in v0.32. (@rfratto)
-
grafana-agent-operator
supports oauth2 as an authentication method for remote_write. (@timo-42) -
Grafana Agent Flow: Add tracing instrumentation and a
tracing
block to forward traces tootelcol
component. (@rfratto) -
Grafana Agent Flow: Add a
discovery_target_decode
function to decode a JSON array of discovery targets corresponding to Prometheus' HTTP and file service discovery formats. (@rfratto) -
New Grafana Agent Flow components:
-
remote.http
polls an HTTP URL and exposes the response body as a string or secret to other components. (@rfratto) -
discovery.docker
discovers Docker containers from a Docker Engine host. (@rfratto) -
loki.source.file
reads and tails files for log entries and forwards them to otherloki
components. (@tpaschalis) -
loki.write
receives log entries from otherloki
components and sends them over to a Loki instance. (@tpaschalis) -
loki.relabel
receives log entries from otherloki
components and rewrites their label set. (@tpaschalis) -
loki.process
receives log entries from otherloki
components and runs one or more processing stages. (@tpaschalis) -
discovery.file
discovers files on the filesystem following glob patterns. (@mattdurham)
-
-
Integrations: Introduce the
snowflake
integration. (@binaryfissiongames)
-
Update agent-loki.yaml to use environment variables in the configuration file (@go4real)
-
Integrations: Always use direct connection in mongodb_exporter integration. (@v-zhuravlev)
-
Update OpenTelemetry Collector dependency to v0.63.1. (@tpaschalis)
-
riverfmt: Permit empty blocks with both curly braces on the same line. (@rfratto)
-
riverfmt: Allow function arguments to persist across different lines. (@rfratto)
-
Flow: The HTTP server will now start before the Flow controller performs the initial load. This allows metrics and pprof data to be collected during the first load. (@rfratto)
-
Add support for using a password map file in
redis_exporter
. (@spartan0x117) -
Flow: Add support for exemplars in Prometheus component pipelines. (@rfratto)
-
Update Prometheus dependency to v2.40.5. (@rfratto)
-
Update Promtail dependency to k127. (@rfratto)
-
Native histograms are now supported in the static Grafana Agent and in
prometheus.*
Flow components. Native histograms will be automatically collected from supported targets. remote_write must be configured to forward native histograms from the WAL to the specified endpoints. (@rfratto) -
Flow: metrics generated by upstream OpenTelemetry Collector components are now exposed at the
/metrics
endpoint of Grafana Agent Flow. (@rfratto)
-
Fix issue where whitespace was being sent as part of password when using a password file for
redis_exporter
. (@spartan0x117) -
Flow UI: Fix issue where a configuration block referencing a component would cause the graph page to fail to load. (@rfratto)
-
Remove duplicate
oauth2
key frommetricsinstances
CRD. (@daper) -
Fix issue where on checking whether to restart integrations the Integration Manager was comparing configs with secret values scrubbed, preventing reloads if only secrets were updated. (@spartan0x117)
- Grafana Agent Flow has graduated from experimental to beta.
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
- JSON-encoded traces from OTLP versions earlier than 0.16.0 are no longer supported. (@rfratto)
- The binary names
agent
,agentctl
, andagent-operator
have been deprecated and will be renamed tografana-agent
,grafana-agentctl
, andgrafana-agent-operator
in the v0.31.0 release.
-
Add
agentctl test-logs
command to allow testing log configurations by redirecting collected logs to standard output. This can be useful for debugging. (@jcreixell) -
New Grafana Agent Flow components:
-
otelcol.receiver.otlp
receives OTLP-formatted traces, metrics, and logs. Data can then be forwarded to otherotelcol
components. (@rfratto) -
otelcol.processor.batch
batches data fromotelcol
components before forwarding it to otherotelcol
components. (@rfratto) -
otelcol.exporter.otlp
accepts data fromotelcol
components and sends it to a gRPC server using the OTLP protocol. (@rfratto) -
otelcol.exporter.otlphttp
accepts data fromotelcol
components and sends it to an HTTP server using the OTLP protocol. (@tpaschalis) -
otelcol.auth.basic
performs basic authentication forotelcol
components that support authentication extensions. (@rfratto) -
otelcol.receiver.jeager
receives Jaeger-formatted traces. Data can then be forwarded to otherotelcol
components. (@rfratto) -
otelcol.processor.memory_limiter
periodically checks memory usage and drops data or forces a garbage collection if the defined limits are exceeded. (@tpaschalis) -
otelcol.auth.bearer
performs bearer token authentication forotelcol
components that support authentication extensions. (@rfratto) -
otelcol.auth.headers
attaches custom request headers tootelcol
components that support authentication extensions. (@rfratto) -
otelcol.receiver.prometheus
receives Prometheus metrics, converts them to the OTLP metric format and forwards them to otherotelcol
components. (@tpaschalis) -
otelcol.exporter.prometheus
forwards OTLP-formatted data to compatibleprometheus
components. (@rfratto)
-
-
Flow: Allow config blocks to reference component exports. (@tpaschalis)
-
Introduce
/-/support
endpoint for generating 'support bundles' in static agent mode. Support bundles are zip files of commonly-requested information that can be used to debug a running agent. (@tpaschalis)
-
Update OpenTelemetry Collector dependency to v0.61.0. (@rfratto)
-
Add caching to Prometheus relabel component. (@mattdurham)
-
Grafana Agent Flow: add
agent_resources_*
metrics which explain basic platform-agnostic metrics. These metrics assist with basic monitoring of Grafana Agent, but are not meant to act as a replacement for fully featured components likeprometheus.integration.node_exporter
. (@rfratto) -
Enable field label in TenantStageSpec of PodLogs pipeline. (@siiimooon)
-
Enable reporting of enabled integrations. (@marctc)
-
Grafana Agent Flow:
prometheus.remote_write
andprometheus.relabel
will now export receivers immediately, removing the need for dependant components to be evaluated twice at process startup. (@rfratto) -
Add missing setting to configure instance key for Eventhandler integration. (@marctc)
-
Update Prometheus dependency to v2.39.1. (@rfratto)
-
Update Promtail dependency to weekly release k122. (@rfratto)
-
Tracing: support the
num_traces
andexpected_new_traces_per_sec
configuration parameters in the tail_sampling processor. (@ptodev)
-
Remove empty port from the
apache_http
integration's instance label. (@katepangLiu) -
Fix identifier on target creation for SNMP v2 integration. (@marctc)
-
Fix bug when specifying Blackbox's modules when using Blackbox integration. (@marctc)
-
Tracing: fix a panic when the required
protocols
field was not set in theotlp
receiver. (@ptodev) -
Support Bearer tokens for metric remote writes in the Grafana Operator (@jcreixell, @marctc)
-
Update versions of embedded Prometheus exporters used for integrations:
-
Update
github.com/prometheus/statsd_exporter
tov0.22.8
. (@captncraig) -
Update
github.com/prometheus-community/postgres_exporter
tov0.11.1
. (@captncraig) -
Update
github.com/prometheus/memcached_exporter
tov0.10.0
. (@captncraig) -
Update
github.com/prometheus-community/elasticsearch_exporter
tov1.5.0
. (@captncraig) -
Update
github.com/prometheus/mysqld_exporter
tov0.14.0
. (@captncraig) -
Update
github.com/prometheus/consul_exporter
tov0.8.0
. (@captncraig) -
Update
github.com/ncabatoff/process-exporter
tov0.7.10
. (@captncraig) -
Update
github.com/prometheus-community/postgres_exporter
tov0.11.1
. (@captncraig)
-
-
Use Go 1.19.3 for builds. (@rfratto)
- Update Docker base image to resolve OpenSSL vulnerabilities CVE-2022-3602 and CVE-2022-3786. Grafana Agent does not use OpenSSL, so we do not believe it is vulnerable to these issues, but the base image has been updated to remove the report from image scanners. (@rfratto)
-
Introduce Grafana Agent Flow, an experimental "programmable pipeline" runtime mode which improves how to configure and debug Grafana Agent by using components. (@captncraig, @karengermond, @marctc, @mattdurham, @rfratto, @rlankfo, @tpaschalis)
-
Introduce Blackbox exporter integration. (@marctc)
- Update Loki dependency to v2.6.1. (@rfratto)
-
Fix relabel configs in sample agent-operator manifests (@hjet)
-
Operator no longer set the
SecurityContext.Privileged
flag in theconfig-reloader
container. (@hsyed-dojo) -
Add metrics for config reloads and config hash (@jcreixell)
NOTE: ARMv6 Docker images are no longer being published.
We have stopped publishing Docker images for ARMv6 platforms. This is due to the new Ubuntu base image we are using that does not support ARMv6. The new Ubuntu base image has less reported CVEs, and allows us to provide more secure Docker images. We will still continue to publish ARMv6 release binaries and deb/rpm packages.
- Switch docker image base from debian to ubuntu. (@captncraig)
-
Integrations: (beta) Add vmware_exporter integration (@rlankfo)
-
App agent receiver: add Event kind to payload (@domasx2)
-
Tracing: Introduce a periodic appender to the remotewriteexporter to control sample rate. (@mapno)
-
Tracing: Update OpenTelemetry dependency to v0.55.0. (@rfratto, @mapno)
-
Add base agent-operator jsonnet library and generated manifests (@hjet)
-
Add full (metrics, logs, K8s events) sample agent-operator jsonnet library and gen manifests (@hjet)
-
Introduce new configuration fields for disabling Keep-Alives and setting the IdleConnectionTimeout when scraping. (@tpaschalis)
-
Add field to Operator CRD to disable report usage functionality. (@marctc)
-
Tracing: Fixed issue with the PromSD processor using the
connection
method to discover the IP address. It was failing to match because the port number was included in the address string. (@jphx) -
Register prometheus discovery metrics. (@mattdurham)
-
Fix seg fault when no instance parameter is provided for apache_http integration, using integrations-next feature flag. (@rgeyer)
-
Fix grafanacloud-install.ps1 web request internal server error when fetching config. (@rlankfo)
-
Fix snmp integration not passing module or walk_params parameters when scraping. (@rgeyer)
-
Fix unmarshal errors (key "<walk_param name>" already set in map) for snmp integration config when walk_params is defined, and the config is reloaded. (@rgeyer)
-
Update several go dependencies to resolve warnings from certain security scanning tools. None of the resolved vulnerabilities were known to be exploitable through the agent. (@captncraig)
-
It is now possible to compile Grafana Agent using Go 1.19. (@rfratto)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
- Change windows certificate store so client certificate is no longer required in store. (@mattdurham)
-
Operator: Fix issue where configured
targetPort
ServiceMonitors resulted in generating an incorrect scrape_config. (@rfratto) -
Build the Linux/AMD64 artifacts using the opt-out flag for the ebpf_exporter. (@tpaschalis)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
-
Deprecated
server
YAML block fields have now been removed in favor of the command-line flags that replaced them. These fields were originally deprecated in v0.24.0. (@rfratto) -
Changed tail sampling policies to be configured as in the OpenTelemetry Collector. (@mapno)
-
Introduce Apache HTTP exporter integration. (@v-zhuravlev)
-
Introduce eBPF exporter integration. (@tpaschalis)
- Truncate all records in WAL if repair attempt fails. (@rlankfo)
-
Relative symlinks for promtail now work as expected. (@RangerCD, @mukerjee)
-
Fix rate limiting implementation for the app agent receiver integration. (@domasx2)
-
Fix mongodb exporter so that it now collects all metrics. (@mattdurham)
-
Integer types fail to unmarshal correctly in operator additional scrape configs. (@rlankfo)
-
Unwrap replayWAL error before attempting corruption repair. (@rlankfo)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
- Traces: Use
rpc.grpc.status_code
attribute to determine span failed in the service graph processor (@rcrowe)
-
Add HTTP endpoints to fetch active instances and targets for the Logs subsystem. (@marctc)
-
(beta) Add support for using windows certificate store for TLS connections. (@mattdurham)
-
Grafana Agent Operator: add support for integrations through an
Integration
CRD which is discovered byGrafanaAgent
. (@rfratto) -
(experimental) Add app agent receiver integration. This depends on integrations-next being enabled via the
integrations-next
feature flag. Use-enable-features=integrations-next
to use this integration. (@kpelelis, @domas) -
Introduce SNMP exporter integration. (@v-zhuravlev)
-
Configure the agent to report the use of feature flags to grafana.com. (@marctc)
-
integrations-next: Integrations using autoscrape will now autoscrape metrics using in-memory connections instead of connecting to themselves over the network. As a result of this change, the
client_config
field has been removed. (@rfratto) -
Enable
proxy_url
support onoauth2
for metrics and logs (update prometheus/common dependency tov0.33.0
). (@martin-jaeger-maersk) -
extra-scrape-metrics
can now be enabled with the--enable-features=extra-scrape-metrics
feature flag. See https://prometheus.io/docs/prometheus/2.31/feature_flags/#extra-scrape-metrics for details. (@rlankfo) -
Resolved issue in v2 integrations where if an instance name was a prefix of another the route handler would fail to match requests on the longer name (@mattdurham)
-
Set
include_metadata
to true by default for OTLP traces receivers (@mapno)
- Scraping service was not honoring the new server grpc flags
server.grpc.address
. (@mattdurham)
-
Update base image of official Docker containers from Debian buster to Debian bullseye. (@rfratto)
-
Use Go 1.18 for builds. (@rfratto)
-
Add
metrics
prefix to the url of list instances endpoint (GET /agent/api/v1/instances
) and list targets endpoint (GET /agent/api/v1/metrics/targets
). (@marctc) -
Add extra identifying labels (
job
,instance
,agent_hostname
) to eventhandler integration. (@hjet) -
Add
extra_labels
configuration to eventhandler integration. (@hjet)
- Added configuration watcher delay to prevent race condition in cases where scraping service mode has not gracefully exited. (@mattdurham)
- Update version of node_exporter to include additional metrics for osx. (@v-zhuravlev)
-
Add missing version information back into
agentctl --version
. (@rlankfo) -
Bump version of github-exporter to latest upstream SHA 284088c21e7d, which includes fixes from bugs found in their latest tag. This includes a fix where not all releases where retrieved when pulling release information. (@rfratto)
-
Set the
Content-Type
HTTP header toapplication/json
for API endpoints returning json objects. (@marctc) -
Operator: fix issue where a
username_file
field was incorrectly set. (@rfratto) -
Initialize the logger with default
log_level
andlog_format
parameters. (@tpaschalis)
- Embed timezone data to enable Promtail pipelines using the
location
field on Windows machines. (@tpaschalis)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
GRAFANA AGENT OPERATOR USERS: As of this release, Grafana Agent Operator does not support versions of Grafana Agent prior to v0.24.0.
-
The following metrics will now be prefixed with
agent_dskit_
instead ofcortex_
:cortex_kv_request_duration_seconds
,cortex_member_consul_heartbeats_total
,cortex_member_ring_tokens_owned
,cortex_member_ring_tokens_to_own
,cortex_ring_member_ownership_percent
,cortex_ring_members
,cortex_ring_oldest_member_timestamp
,cortex_ring_tokens_owned
,cortex_ring_tokens_total
. (@rlankfo) -
Traces: the
traces_spanmetrics_calls_total_total
metric has been renamed totraces_spanmetrics_calls_total
(@fredr) -
Two new flags,
-server.http.enable-tls
and-server.grpc.enable-tls
must be provided to explicitly enable TLS support. This is a change of the previous behavior where TLS support was enabled when a certificate pair was provided. (@rfratto) -
Many command line flags starting with
-server.
block have been renamed. (@rfratto) -
The
-log.level
and-log.format
flags are removed in favor of being set in the configuration file. (@rfratto) -
Flags for configuring TLS have been removed in favor of being set in the configuration file. (@rfratto)
-
Dynamic reload is no longer supported for deprecated server block fields. Changing a deprecated field will be ignored and cause the reload to fail. (@rfratto)
-
The default HTTP listen address is now
127.0.0.1:12345
. Use the-server.http.address
flag to change this value. (@rfratto) -
The default gRPC listen address is now
127.0.0.1:12346
. Use the-server.grpc.address
flag to change this value. (@rfratto) -
-reload-addr
and-reload-port
have been removed. They are no longer necessary as the primary HTTP server is now static and can't be shut down in the middle of a/-/reload
call. (@rfratto) -
(Only impacts
integrations-next
feature flag) Many integrations have been renamed to better represent what they are integrating with. For example,redis_exporter
is nowredis
. This change requires updatingintegrations-next
-enabled configuration files. This change also changes integration names shown in metric labels. (@rfratto) -
The deprecated
-prometheus.*
flags have been removed in favor of their-metrics.*
counterparts. The-prometheus.*
flags were first deprecated in v0.19.0. (@rfratto)
- Most fields in the
server
block of the configuration file are now deprecated in favor of command line flags. These fields will be removed in the v0.26.0 release. Please consult the upgrade guide for more information and rationale. (@rfratto)
-
Added config read API support to GrafanaAgent Custom Resource Definition. (@shamsalmon)
-
Added consulagent_sd to target discovery. (@chuckyz)
-
Introduce EXPERIMENTAL support for dynamic configuration. (@mattdurham)
-
Introduced endpoint that accepts remote_write requests and pushes metrics data directly into an instance's WAL. (@tpaschalis)
-
Added builds for linux/ppc64le. (@aklyachkin)
-
Tracing: Exporters can now be configured to use OAuth. (@canuteson)
-
Strengthen readiness check for metrics instances. (@tpaschalis)
-
Parameterize namespace field in sample K8s logs manifests (@hjet)
-
Upgrade to Loki k87. (@rlankfo)
-
Update Prometheus dependency to v2.34.0. (@rfratto)
-
Update OpenTelemetry-collector dependency to v0.46.0. (@mapno)
-
Update cAdvisor dependency to v0.44.0. (@rfratto)
-
Update mongodb_exporter dependency to v0.31.2 (@mukerjee)
-
Use grafana-agent/v2 Tanka Jsonnet to generate K8s manifests (@hjet)
-
Replace agent-bare.yaml K8s sample Deployment with StatefulSet (@hjet)
-
Improve error message for
agentctl
when timeout happens callingcloud-config
command (@marctc) -
Enable integrations-next by default in agent-bare.yaml. Please note #1262 (@hjet)
-
Fix Kubernetes manifests to use port
4317
for OTLP instead of the previous55680
in line with the default exposed port in the agent. -
Ensure singleton integrations are honored in v2 integrations (@mattdurham)
-
Tracing:
const_labels
is now correctly parsed in the remote write exporter. (@fredr) -
integrations-next: Fix race condition where metrics endpoints for integrations may disappear after reloading the config file. (@rfratto)
-
Removed the
server.path_prefix
field which would break various features in Grafana Agent when set. (@rfratto) -
Fix issue where installing the DEB/RPM packages would overwrite the existing config files and environment files. (@rfratto)
-
Set
grafanaDashboardFolder
as top level key in the mixin. (@Duologic) -
Operator: Custom Secrets or ConfigMaps to mount will no longer collide with the path name of the default secret mount. As a side effect of this bugfix, custom Secrets will now be mounted at
/var/lib/grafana-agent/extra-secrets/<secret name>
and custom ConfigMaps will now be mounted at/var/lib/grafana-agent/extra-configmaps/<configmap name>
. This is not a breaking change as it was previously impossible to properly provide these custom mounts. (@rfratto) -
Flags accidentally prefixed with
-metrics.service..
(two.
in a row) have now been fixed to only have one.
. (@rfratto) -
Protect concurrent writes to the WAL in the remote write exporter (@mapno)
- The
-metrics.wal-directory
flag andmetrics.wal_directory
config option will now default todata-agent/
, the same default WAL directory as Prometheus Agent. (@rfratto)
-
Go 1.17 is now used for all builds of the Agent. (@tpaschalis)
-
integrations-next: Add
extra_labels
to add a custom set of labels to integration targets. (@rfratto) -
The agent no longer appends duplicate exemplars. (@tpaschalis)
-
Added Kubernetes eventhandler integration (@hjet)
-
Enables sending of exemplars over remote write by default. (@rlankfo)
-
Fixed issue where Grafana Agent may panic if there is a very large WAL loading while old WALs are being deleted or the
/agent/api/v1/targets
endpoint is called. (@tpaschalis) -
Fix panic in prom_sd_processor when address is empty (@mapno)
-
Operator: Add missing proxy_url field from generated remote_write configs. (@rfratto)
-
Honor the specified log format in the traces subsystem (@mapno)
-
Fix typo in node_exporter for runit_service_dir. (@mattdurham)
-
Allow inlining credentials in remote_write url. (@tpaschalis)
-
integrations-next: Wait for integrations to stop when starting new instances or shutting down (@rfratto).
-
Fix issue with windows_exporter mssql collector crashing the agent. (@mattdurham)
-
The deb and rpm files will now ensure the /var/lib/grafana-agent data directory is created with permissions set to 0770. (@rfratto)
-
Make agent-traces.yaml Namespace a template-friendly variable (@hjet)
-
Disable
machine-id
journal vol by default in sample logs manifest (@hjet)
This release has deprecations. Please read entries carefully and consult the upgrade guide for specific instructions.
-
The node_exporter integration's
netdev_device_whitelist
field is deprecated in favor ofnetdev_device_include
. Support for the old field name will be removed in a future version. (@rfratto) -
The node_exporter integration's
netdev_device_blacklist
field is deprecated in favor ofnetdev_device_include
. Support for the old field name will be removed in a future version. (@rfratto) -
The node_exporter integration's
systemd_unit_whitelist
field is deprecated in favor ofsystemd_unit_include
. Support for the old field name will be removed in a future version. (@rfratto) -
The node_exporter integration's
systemd_unit_blacklist
field is deprecated in favor ofsystemd_unit_exclude
. Support for the old field name will be removed in a future version. (@rfratto) -
The node_exporter integration's
filesystem_ignored_mount_points
field is deprecated in favor offilesystem_mount_points_exclude
. Support for the old field name will be removed in a future version. (@rfratto) -
The node_exporter integration's
filesystem_ignored_fs_types
field is deprecated in favor offilesystem_fs_types_exclude
. Support for the old field name will be removed in a future version. (@rfratto)
-
(beta) Enable experimental config urls for fetching remote configs. Currently, only HTTP/S is supported. Pass the
-enable-features=remote-configs
flag to turn this on. (@rlankfo) -
Added cAdvisor integration. (@rgeyer)
-
Traces: Add
Agent Tracing Pipeline
dashboard and alerts (@mapno) -
Traces: Support jaeger/grpc exporter (@nicoche)
-
(beta) Enable an experimental integrations subsystem revamp. Pass
integrations-next
to-enable-features
to turn this on. Reading the documentation for the revamp is recommended; enabling it causes breaking config changes. (@rfratto)
-
Traces: Improved pod association in PromSD processor (@mapno)
-
Updated OTel to v0.40.0 (@mapno)
-
Remote write dashboard: show in and out sample rates (@bboreham)
-
Remote write dashboard: add mean latency (@bboreham)
-
Update node_exporter dependency to v1.3.1. (@rfratto)
-
Cherry-pick Prometheus PR #10102 into our Prometheus dependency (@rfratto).
-
Fix usage of POSTGRES_EXPORTER_DATA_SOURCE_NAME when using postgres_exporter integration (@f11r)
-
Change ordering of the entrypoint for windows service so that it accepts commands immediately (@mattdurham)
-
Only stop WAL cleaner when it has been started (@56quarters)
-
Fix issue with unquoted install path on Windows, that could allow escalation or running an arbitrary executable (@mattdurham)
-
Fix cAdvisor so it collects all defined metrics instead of the last (@pkoenig10)
-
Fix panic when using 'stdout' in automatic logging (@mapno)
-
Grafana Agent Operator: The /-/ready and /-/healthy endpoints will no longer always return 404 (@rfratto).
- Remove log-level flag from systemd unit file (@jpkrohling)
- This release contains a fix for CVE-2021-41090.
- This release disables the existing
/-/config
and/agent/api/v1/configs/{name}
endpoints by default. Pass the--config.enable-read-api
flag at the command line to opt in to these endpoints.
-
Fix panic when using postgres_exporter integration (@saputradharma)
-
Fix panic when dnsamsq_exporter integration tried to log a warning (@rfratto)
-
Statsd Integration: Adding logger instance to the statsd mapper instantiation. (@gaantunes)
-
Statsd Integration: Fix issue where mapped metrics weren't exposed to the integration. (@mattdurham)
-
Operator: fix bug where version was a required field (@rfratto)
-
Metrics: Only run WAL cleaner when metrics are being used and a WAL is configured. (@rfratto)
-
Update Cortex dependency to v1.10.0-92-g85c378182. (@rlankfo)
-
Update Loki dependency to v2.1.0-656-g0ae0d4da1. (@rlankfo)
-
Update Prometheus dependency to v2.31.0 (@rlankfo)
-
Add Agent Operator Helm quickstart guide (@hjet)
-
Reorg Agent Operator quickstart guides (@hjet)
-
Packaging: Use correct user/group env variables in RPM %post script (@simonc6372)
-
Validate logs config when using logs_instance with automatic logging processor (@mapno)
-
Operator: Fix MetricsInstance Service port (@hjet)
-
Operator: Create govern service per Grafana Agent (@shturman)
-
Operator: Fix relabel_config directive for PodLogs resource (@hjet)
-
Traces: Fix
success_logic
code in service graphs processor (@mapno)
-
Self-scraped integrations will now use an SUO-specific value for the
instance
label. (@rfratto) -
Traces: Changed service graphs store implementation to improve CPU performance (@mapno)
NOTE: The fixes in this patch are only present in v0.20.1 and >=v0.21.2.
- This release contains a fix for CVE-2021-41090.
- This release disables the existing
/-/config
and/agent/api/v1/configs/{name}
endpoitns by default. Pass the--config.enable-read-api
flag at the command line to opt in to these endpoints.
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
- push_config is no longer supported in trace's config (@mapno)
-
Operator: The Grafana Agent Operator can now generate a Kubelet service to allow a ServiceMonitor to collect Kubelet and cAdvisor metrics. This requires passing a
--kubelet-service
flag to the Operator innamespace/name
format (likekube-system/kubelet
). (@rfratto) -
Service graphs processor (@mapno)
-
Updated mysqld_exporter to v0.13.0 (@gaantunes)
-
Updated postgres_exporter to v0.10.0 (@gaantunes)
-
Updated redis_exporter to v1.27.1 (@gaantunes)
-
Updated memcached_exporter to v0.9.0 (@gaantunes)
-
Updated statsd_exporter to v0.22.2 (@gaantunes)
-
Updated elasticsearch_exporter to v1.2.1 (@gaantunes)
-
Add remote write to silent Windows Installer (@mattdurham)
-
Updated mongodb_exporter to v0.20.7 (@rfratto)
-
Updated OTel to v0.36 (@mapno)
-
Updated statsd_exporter to v0.22.2 (@mattdurham)
-
Update windows_exporter to v0.16.0 (@rfratto, @mattdurham)
-
Add send latency to agent dashboard (@bboreham)
-
Do not immediately cancel context when creating a new trace processor. This was preventing scrape_configs in traces from functioning. (@lheinlen)
-
Sanitize autologged Loki labels by replacing invalid characters with underscores (@mapno)
-
Traces: remove extra line feed/spaces/tabs when reading password_file content (@nicoche)
-
Updated envsubst to v2.0.0-20210730161058-179042472c46. This version has a fix needed for escaping values outside of variable substitutions. (@rlankfo)
-
Grafana Agent Operator should no longer delete resources matching the names of the resources it manages. (@rfratto)
-
Grafana Agent Operator will now appropriately assign an
app.kubernetes.io/managed-by=grafana-agent-operator
to all created resources. (@rfratto)
-
Configuration API now returns 404 instead of 400 when attempting to get or delete a config which does not exist. (@kgeckhart)
-
The windows_exporter now disables the textfile collector by default. (@rfratto)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
-
Reduced verbosity of tracing autologging by not logging
STATUS_CODE_UNSET
status codes. (@mapno) -
Operator: rename
Prometheus*
CRDs toMetrics*
andPrometheus*
fields toMetrics*
. (@rfratto) -
Operator: CRDs are no longer referenced using a hyphen in the name to be consistent with how Kubernetes refers to resources. (@rfratto)
-
prom_instance
in the spanmetrics config is now namedmetrics_instance
. (@rfratto)
-
The
loki
key at the root of the config file has been deprecated in favor oflogs
.loki
-named fields inautomatic_logging
have been renamed accordinly:loki_name
is nowlogs_instance_name
,loki_tag
is nowlogs_instance_tag
, andbackend: loki
is nowbackend: logs_instance
. (@rfratto) -
The
prometheus
key at the root of the config file has been deprecated in favor ofmetrics
. Flag names starting withprometheus.
have also been deprecated in favor of the same flags with themetrics.
prefix. Metrics prefixed withagent_prometheus_
are now prefixed withagent_metrics_
. (@rfratto) -
The
tempo
key at the root of the config file has been deprecated in favor oftraces
. (@mattdurham)
-
Added GitHub exporter integration. (@rgeyer)
-
Add TLS config options for tempo
remote_write
s. (@mapno) -
Support autologging span attributes as log labels (@mapno)
-
Put Tests requiring Network Access behind a -online flag (@flokli)
-
Add logging support to the Grafana Agent Operator. (@rfratto)
-
Add
operator-detach
command to agentctl to allow zero-downtime upgrades when removing an Operator CRD. (@rfratto) -
The Grafana Agent Operator will now default to deploying the matching release version of the Grafana Agent instead of v0.14.0. (@rfratto)
-
Update OTel dependency to v0.30.0 (@mapno)
-
Allow reloading configuration using
SIGHUP
signal. (@tharun208) -
Add HOSTNAME environment variable to service file to allow for expanding the $HOSTNAME variable in agent config. (@dfrankel33)
-
Update jsonnet-libs to 1.21 for Kubernetes 1.21+ compatability. (@MurzNN)
-
Make method used to add k/v to spans in prom_sd processor configurable. (@mapno)
-
Regex capture groups like
${1}
will now be kept intact when using-config.expand-env
. (@rfratto) -
The directory of the logs positions file will now properly be created on startup for all instances. (@rfratto)
-
The Linux system packages will now configure the grafana-agent user to be a member of the adm and systemd-journal groups. This will allow logs to read from journald and /var/log by default. (@rfratto)
-
Fix collecting filesystem metrics on Mac OS (darwin) in the
node_exporter
integration default config. (@eamonryan) -
Remove v0.0.0 flags during build with no explicit release tag (@mattdurham)
-
Fix issue with global scrape_interval changes not reloading integrations (@kgeckhart)
-
Grafana Agent Operator will now detect changes to referenced ConfigMaps and Secrets and reload the Agent properly. (@rfratto)
-
Grafana Agent Operator's object label selectors will now use Kubernetes defaults when undefined (i.e., default to nothing). (@rfratto)
-
Fix yaml marshalling tag for cert_file in kafka exporter agent config. (@rgeyer)
-
Fix warn-level logging of dropped targets. (@james-callahan)
-
Standardize scrape_interval to 1m in examples. (@mattdurham)
- Add
agent_prometheus_configs_changed_total
metric to track instance config events. (@rfratto)
-
Fix info logging on windows. (@mattdurham)
-
Scraping service: Ensure that a reshard is scheduled every reshard interval. (@rfratto)
-
Register missing metric for configstore consul request duration. (@rfratto)
-
Logs should contain a caller field with file and line numbers again (@kgeckhart)
-
In scraping service mode, the polling configuration refresh should honor timeout. (@mattdurham)
-
In scraping service mode, the lifecycle reshard should happen using a goroutine. (@mattdurham)
-
In scraping service mode, scraping service can deadlock when reloading during join. (@mattdurham)
-
Scraping service: prevent more than one refresh from being queued at a time. (@rfratto)
- Honor the prefix and remove prefix from consul list results (@mattdurham)
- Reduce number of consul calls when ran in scrape service mode (@mattdurham)
-
Added GitHub exporter integration. (@rgeyer)
-
Add support for OTLP HTTP trace exporting. (@mapno)
-
Switch to drone for releases. (@mattdurham)
-
Update postgres_exporter to a branch of v0.10.0
- Enabled flag for integrations is not being honored. (@mattdurham)
- Added Kafka Lag exporter integration. (@gaantunes)
- Fix race condition that may occur and result in a panic when initializing scraping service cluster. (@rfratto)
- Fix issue where replaying a WAL caused incorrect metrics to be sent over remote write. (@rfratto)
- (beta) A Grafana Agent Operator is now available. (@rfratto)
- Error messages when installing the Grafana Agent for Grafana Cloud will now be shown. (@rfratto)
-
Fix a leak in the shared string interner introduced in v0.14.0. This fix was made to a dependency. (@rfratto)
-
Fix issue where a target will fail to be scraped for the process lifetime if that target had gone down for long enough that its series were removed from the in-memory cache (2 GC cycles). (@rfratto)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
- The configuration of Tempo Autologging has changed. (@mapno)
- Add support for exemplars. (@mapno)
-
Add the option to log to stdout instead of a Loki instance. (@joe-elliott)
-
Update Cortex dependency to v1.8.0.
-
Running the Agent as a DaemonSet with host_filter and role: pod should no longer cause unnecessary load against the Kubernetes SD API. (@rfratto)
-
Update Prometheus to v2.27.0. (@mapno)
-
Update Loki dependency to d88f3996eaa2. This is a non-release build, and was needed to support exemplars. (@mapno)
-
Update Cortex dependency to d382e1d80eaf. This is a non-release build, and was needed to support exemplars. (@mapno)
-
Host filter relabeling rules should now work. (@rfratto)
-
Fixed issue where span metrics where being reported with wrong time unit. (@mapno)
- Intentionally order tracing processors. (@joe-elliott)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
STABILITY NOTICE: As of this release, functionality that is not recommended for production use and is expected to change will be tagged interchangably as "experimental" or "beta."
- The Scraping service API will now reject configs that read credentials from
disk by default. This prevents malicious users from reading arbitrary files
and sending their contents over the network. The old behavior can be
re-enabled by setting
dangerous_allow_reading_files: true
in the scraping service config. (@rfratto)
- Configuration for SigV4 has changed. (@rfratto)
push_config
is now supplanted byremote_block
andbatch
.push_config
will be removed in a future version (@mapno)
-
(beta) New integration: windows_exporter (@mattdurham)
-
(beta) Grafana Agent Windows Installer is now included as a release artifact. (@mattdurham)
-
Official M1 Mac release builds will now be generated! Look for
agent-darwin-arm64
andagentctl-darwin-arm64
in the release assets. (@rfratto) -
Add support for running as a Windows service (@mattdurham)
-
(beta) Add /-/reload support. It is not recommended to invoke
/-/reload
against the main HTTP server. Instead, two new command-line flags have been added:--reload-addr
and--reload-port
. These will launch a/-/reload
-only HTTP server that can be used to safely reload the Agent's state. (@rfratto) -
Add a /-/config endpoint. This endpoint will return the current configuration file with defaults applied that the Agent has loaded from disk. (@rfratto)
-
(beta) Support generating metrics and exposing them via a Prometheus exporter from span data. (@yeya24)
-
Tail-based sampling for tracing pipelines (@mapno)
-
Added Automatic Logging feature for Tempo (@joe-elliott)
-
Disallow reading files from within scraping service configs by default. (@rfratto)
-
Add remote write for span metrics (@mapno)
-
Support compression for trace export. (@mdisibio)
-
Add global remote_write configuration that is shared between all instances and integrations. (@mattdurham)
-
Go 1.16 is now used for all builds of the Agent. (@rfratto)
-
Update Prometheus dependency to v2.26.0. (@rfratto)
-
Upgrade
go.opentelemetry.io/collector
to v0.21.0 (@mapno) -
Add kafka trace receiver (@mapno)
-
Support mirroring a trace pipeline to multiple backends (@mapno)
-
Add
headers
field inremote_write
config for Tempo.headers
specifies HTTP headers to forward to the remote endpoint. (@alexbiehl) -
Add silent uninstall to Windows Uninstaller. (@mattdurham)
-
Native Darwin arm64 builds will no longer crash when writing metrics to the WAL. (@rfratto)
-
Remote write endpoints that never function across the lifetime of the Agent will no longer prevent the WAL from being truncated. (@rfratto)
-
Bring back FreeBSD support. (@rfratto)
-
agentctl will no longer leak WAL resources when retrieving WAL stats. (@rfratto)
-
Ensure defaults are applied to undefined sections in config file. This fixes a problem where integrations didn't work if
prometheus:
wasn't configured. (@rfratto) -
Fixed issue where automatic logging double logged "svc". (@joe-elliott)
-
The Grafana Cloud Agent has been renamed to the Grafana Agent. (@rfratto)
-
Instance configs uploaded to the Config Store API will no longer be stored along with the global Prometheus defaults. This is done to allow globals to be updated and re-apply the new global defaults to the configs from the Config Store. (@rfratto)
-
The User-Agent header sent for logs will now be
GrafanaAgent/<version>
(@rfratto) -
Add
tempo_spanmetrics
namespace in spanmetrics (@mapno)
- Validate that incoming scraped metrics do not have an empty label set or a label set with duplicate labels, mirroring the behavior of Prometheus. (@rfratto)
The primary branch name has changed from
master
tomain
. You may have to update your local checkouts of the repository to point at the new branch name.
- postgres_exporter: Support query_path and disable_default_metrics. (@rfratto)
-
Support other architectures in installation script. (@rfratto)
-
Allow specifying custom wal_truncate_frequency per integration. (@rfratto)
-
The SigV4 region can now be inferred using the shared config (at
$HOME/.aws/config
) or environment variables (viaAWS_CONFIG
). (@rfratto) -
Update Prometheus dependency to v2.25.0. (@rfratto)
-
Not providing an
-addr
flag foragentctl config-sync
will no longer report an error and will instead use the pre-existing default value. (@rfratto) -
Fixed a bug from v0.12.0 where the Loki installation script failed because positions_directory was not set. (@rfratto)
-
Reduce the likelihood of dataloss during a remote_write-side outage by increasing the default wal_truncation_frequency to 60m and preventing the WAL from being truncated if the last truncation timestamp hasn't changed. This change increases the size of the WAL on average, and users may configure a lower wal_truncation_frequency to deliberately choose a smaller WAL over write guarantees. (@rfratto)
-
Add the ability to read and serve HTTPS integration metrics when given a set certificates (@mattdurham)
BREAKING CHANGES: This release has breaking changes. Please read entries carefully and consult the upgrade guide for specific instructions.
-
The configuration format for the
loki
block has changed. (@rfratto) -
The configuration format for the
tempo
block has changed. (@rfratto)
-
Support for multiple Loki Promtail instances has been added. (@rfratto)
-
Support for multiple Tempo instances has been added. (@rfratto)
-
Added ElasticSearch exporter integration. (@colega)
-
.deb
and.rpm
packages are now generated for all supported architectures. The architecture of the AMD64 package in the filename has been renamed toamd64
to stay synchronized with the architecture name presented from other release assets. (@rfratto) -
The
/agent/api/v1/targets
API will now include discovered labels on the target pre-relabeling in adiscovered_labels
field. (@rfratto) -
Update Loki to 59a34f9867ce. This is a non-release build, and was needed to support multiple Loki instances. (@rfratto)
-
Scraping service: Unhealthy Agents in the ring will no longer cause job distribution to fail. (@rfratto)
-
Scraping service: Cortex ring metrics (prefixed with cortex_ring_) will now be registered for tracking the state of the hash ring. (@rfratto)
-
Scraping service: instance config ownership is now determined by the hash of the instance config name instead of the entire config. This means that updating a config is guaranteed to always hash to the same Agent, reducing the number of metrics gaps. (@rfratto)
-
Only keep a handful of K8s API server metrics by default to reduce default active series usage. (@hjet)
-
Go 1.15.8 is now used for all distributions of the Agent. (@rfratto)
agentctl config-check
will now work correctly when the supplied config file contains integrations. (@hoenn)
-
ARMv6 builds of
agent
andagentctl
will now be included in releases to expand Agent support to cover all models of Raspberry Pis. ARMv6 docker builds are also now available. (@rfratto) -
Added
config-check
subcommand foragentctl
that can be used to validate Agent configuration files before attempting to load them in theagent
itself. (@56quarters)
-
A sigv4 install script for Prometheus has been added. (@rfratto)
-
NAMESPACE may be passed as an environment variable to the Kubernetes install scripts to specify an installation namespace. (@rfratto)
-
The K8s API server scrape job will use the API server Service name when resolving IP addresses for Prometheus service discovery using the "Endpoints" role. (@hjet)
-
The K8s manifests will no longer include the
default/kubernetes
job twice in both the DaemonSet and the Deployment. (@rfratto)
- Prometheus
remote_write
now supports SigV4 authentication using the AWS default credentials chain. This enables the Agent to send metrics to Amazon Managed Prometheus without needing the SigV4 Proxy. (@rfratto)
-
Update
redis_exporter
to v1.15.0. (@rfratto) -
memcached_exporter
has been updated to v0.8.0. (@rfratto) -
process-exporter
has been updated to v0.7.5. (@rfratto) -
wal_cleanup_age
andwal_cleanup_period
have been added to the top-level Prometheus configuration section. These settings control how Write Ahead Logs (WALs) that are not associated with any instances are cleaned up. By default, WALs not associated with an instance that have not been written in the last 12 hours are eligible to be cleaned up. This cleanup can be disabled by settingwal_cleanup_period
to0
. (@56quarters) -
Configuring logs to read from the systemd journal should now work on journals that use +ZSTD compression. (@rfratto)
-
Integrations will now function if the HTTP listen address was set to a value other than the default. (@mattdurham)
-
The default Loki installation will now be able to write its positions file. This was prevented by accidentally writing to a readonly volume mount. (@rfratto)
- agentctl will now be installed by the rpm and deb packages as
grafana-agentctl
. (@rfratto)
-
Add support to configure TLS config for the Tempo exporter to use insecure_skip_verify to disable TLS chain verification. (@bombsimon)
-
Add
sample-stats
toagentctl
to search the WAL and return a summary of samples of series matching the given label selector. (@simonswine) -
New integration: postgres_exporter (@rfratto)
-
New integration: statsd_exporter (@rfratto)
-
New integration: consul_exporter (@rfratto)
-
Add optional environment variable substitution of configuration file. (@dcseifert)
-
min_wal_time
andmax_wal_time
have been added to the instance config settings, guaranteeing that data in the WAL will exist for at leastmin_wal_time
and will not exist for longer thanmax_wal_time
. This change will increase the size of the WAL slightly but will prevent certain scenarios where data is deleted before it is sent. To revert back to the old behavior, setmin_wal_time
to0s
. (@rfratto) -
Update
redis_exporter
to v1.13.1. (@rfratto) -
Bump OpenTelemetry-collector dependency to v0.16.0. (@bombsimon)
-
Fix issue where the Tempo example manifest could not be applied because the port names were too long. (@rfratto)
-
Fix issue where the Agent Kubernetes manifests may not load properly on AKS. (#279) (@rfratto)
- The User-Agent header sent for logs will now be
GrafanaCloudAgent/<version>
(@rfratto)
-
New integration: dnsamsq_exporter (@rfratto).
-
New integration: memcached_exporter (@rfratto).
-
Add
<integration name>_build_info
metric to all integrations. The build info displayed will match the build information of the Agent and not the embedded exporter. This metric is used by community dashboards, so adding it to the Agent increases compatibility with existing dashboards that depend on it existing. (@rfratto) -
Bump OpenTelemetry-collector dependency to 0.14.0 (@joe-elliott)
- Error messages when retrieving configs from the KV store will now be logged, rather than just logging a generic message saying that retrieving the config has failed. (@rfratto)
-
Bump Prometheus dependency to 2.21. (@rfratto)
-
Bump OpenTelemetry-collector dependency to 0.13.0 (@rfratto)
-
Bump Promtail dependency to 2.0. (@rfratto)
-
Enhance host_filtering mode to support targets from Docker Swarm and Consul. Also, add a
host_filter_relabel_configs
to that will apply relabeling rules for determining if a target should be dropped. Add a documentation section explaining all of this in detail. (@rfratto)
-
Fix deb package prerm script so that it stops the agent on package removal. (@jdbaldry)
-
Fix issue where the
push_config
for Tempo field was expected to beremote_write
.push_config
now works as expected. (@rfratto)
- Fix issue where ARM binaries were not published with the GitHub release.
-
Added Tracing Support. (@joe-elliott)
-
Add RPM and deb packaging. (@jdbaldry, @simon6372)
-
arm64 and arm/v7 Docker containers and release builds are now available for
agent
andagentctl
. (@rfratto) -
Add
wal-stats
andtarget-stats
tooling toagentctl
to discover WAL and cardinality issues. (@rfratto) -
mysqld_exporter is now embedded and available as an integration. (@rfratto)
-
redis_exporter is now embedded and available as an integration. (@dafydd-t)
- Resharding the cluster when using the scraping service mode now supports
timeouts through
reshard_timeout
. The default value is30s.
This timeout applies to cluster-wide reshards (performed when joining and leaving the cluster) and local reshards (done on thereshard_interval
). (@rfratto)
-
Fix issue where integrations crashed with instance_mode was set to
distinct
(@rfratto) -
Fix issue where the
agent
integration did not work on Windows (@rfratto). -
Support URL-encoded paths in the scraping service API. (@rfratto)
-
The instance label written from replace_instance_label can now be overwritten with relabel_configs. This bugfix slightly modifies the behavior of what data is stored. The final instance label will now be stored in the WAL rather than computed by remote_write. This change should not negatively affect existing users. (@rfratto)
-
Fix issue where build information was empty when running the Agent with --version. (@rfratto)
-
Fix issue where updating a config in the scraping service may fail to pick up new targets. (@rfratto)
-
Fix deadlock that slowly prevents the Agent from scraping targets at a high scrape volume. (@rfratto)
- The Configs API will now disallow two instance configs having multiple
scrape_configs
with the samejob_name
. This was needed for the instance sharing mode, where combined instances may have duplicatejob_names
across theirscrape_configs
. This brings the scraping service more in line with Prometheus, wherejob_names
must globally be unique. This change also disallows concurrent requests to the put/apply config API endpoint to prevent a race condition of two conflicting configs being applied at the same time. (@rfratto)
use_hostname_label
is now supplanted byreplace_instance_label
.use_hostname_label
will be removed in a future version. (@rfratto)
-
The Grafana Agent can now collect logs and send to Loki. This is done by embedding Promtail, the official Loki log collection client. (@rfratto)
-
Integrations can now be enabled without scraping. Set scrape_integrations to
false
at theintegrations
key or within the specific integration you don't want to scrape. This is useful when another Agent or Prometheus server will scrape the integration. (@rfratto) -
process-exporter is now embedded as
process_exporter
. The hypen has been changed to an underscore in the config file to retain consistency withnode_exporter
. (@rfratto)
-
A new config option,
replace_instance_label
, is now available for use with integrations. When this is true, the instance label for all metrics coming from an integration will be replaced with the machine's hostname rather than 127.0.0.1. (@rfratto) -
The embedded Prometheus version has been updated to 2.20.1. (@rfratto, @gotjosh)
-
The User-Agent header written by the Agent when remote_writing will now be
GrafanaCloudAgent/<Version>
instead ofPrometheus/<Prometheus Version>
. (@rfratto) -
The subsystems of the Agent (
prometheus
,loki
) are now made optional. Enabling integrations also implicitly enables the associated subsystem. For example, enabling theagent
ornode_exporter
integration will force theprometheus
subsystem to be enabled. (@rfratto)
-
The documentation for Tanka configs is now correct. (@amckinley)
-
Minor corrections and spelling issues have been fixed in the Overview documentation. (@amckinley)
-
The new default of
shared
instances mode broke the metric value foragent_prometheus_active_configs
, which was tracking the number of combined configs (i.e., number of launched instances). This metric has been fixed and a new metric,agent_prometheus_active_instances
, has been added to track the numbger of launched instances. If instance sharing is not enabled, both metrics will share the same value. (@rfratto) -
remote_write
names in a group will no longer be copied from the remote_write names of the first instance in the group. Rather, all remote_write names will be generated based on the first 6 characters of the group hash and the first six characters of the remote_write hash. (@rfratto) -
Fix a panic that may occur during shutdown if the WAL is closed in the middle of the WAL being truncated. (@rfratto)
-
A scrape targets API has been added to show every target the Agent is currently scraping, when it was last scraped, how long it took to scrape, and errors from the last scrape, if any. (@rfratto)
-
"Shared Instance Mode" is the new default mode for spawning Prometheus instances, and will improve CPU and memory usage for users of integrations and the scraping service. (@rfratto)
-
Memory stability and utilization of the WAL has been improved, and the reported number of active series in the WAL will stop double-counting recently churned series. (@rfratto)
-
Changing scrape_configs and remote_write configs for an instance will now be dynamically applied without restarting the instance. This will result in less missing metrics for users of the scraping service that change a config. (@rfratto)
-
The Tanka configuration now uses k8s-alpha. (@duologic)
-
The Tanka configuration will now also deploy a single-replica deployment specifically for scraping the Kubernetes API. This deployment acts together with the Daemonset to scrape the full cluster and the control plane. (@gotjosh)
-
The node_exporter filesystem collector will now work on Linux systems without needing to manually set the blocklist and allowlist of filesystems. (@rfratto)
-
Support for integrations has been added. Integrations can be any embedded tool, but are currently used for embedding exporters and generating scrape configs. (@rfratto)
-
node_exporter has been added as an integration. This is the full version of node_exporter with the same configuration options. (@rfratto)
-
An Agent integration that makes the Agent automatically scrape itself has been added. (@rfratto)
- The WAL can now be truncated if running the Agent without any remote_write endpoints. (@rfratto)
-
Prevent the Agent from crashing when a global Prometheus config stanza is not provided. (@robx)
-
Enable agent host_filter in the Tanka configs, which was disabled by default by mistake. (@rfratto)
-
Tanka configs that deploy the scraping service mode are now available (@rfratto)
-
A k3d example has been added as a counterpart to the docker-compose example. (@rfratto)
-
Labels provided by the default deployment of the Agent (Kubernetes and Tanka) have been changed to align with the latest changes to grafana/jsonnet-libs. The old
instance
label is now calledpod
, and the newinstance
label is unique. Acontainer
label has also been added. The Agent mixin has been subsequently updated to also incorporate these label changes. (@rfratto) -
The
remote_write
andscrape_config
sections now share the same validations as Prometheus (@rfratto) -
Setting
wal_truncation_frequency
to less than the scrape interval is now disallowed (@rfratto)
-
A deadlock in scraping service mode when updating a config that shards to the same node has been fixed (@rfratto)
-
remote_write
config stanzas will no longer ignorepassword_file
(@rfratto) -
scrape_config
client secrets (e.g., basic auth, bearer token,password_file
) will now be properly retained in scraping service mode (@rfratto) -
Labels for CPU, RX, and TX graphs in the Agent Operational dashboard now correctly show the pod name of the Agent instead of the exporter name. (@rfratto)
- The Agent has upgraded its vendored Prometheus to v2.18.1 (@gotjosh, @rfratto)
-
A typo in the Tanka configs and Kubernetes manifests that prevents the Agent launching with v0.3.0 has been fixed (@captncraig)
-
Fixed a bug where Tanka mixins could not be used due to an issue with the folder placement enhancement (@rfratto)
agentctl
and the config API will now validate that the YAML they receive are valid instance configs. (@rfratto)
-
A third operational mode called "scraping service mode" has been added. A KV store is used to store instance configs which are distributed amongst a clustered set of Agent processes, dividing the total scrape load across each agent. An API is exposed on the Agents to list, create, update, and delete instance configurations from the KV store. (@rfratto)
-
An "agentctl" binary has been released to interact with the new instance config management API created by the "scraping service mode." (@rfratto, @hoenn)
-
The Agent now includes readiness and healthiness endpoints. (@rfratto)
-
The YAML files are now parsed strictly and an invalid YAML will generate an error at runtime. (@hoenn)
-
The default build mode for the Docker containers is now release, not debug. (@rfratto)
-
The Grafana Agent Tanka Mixins now are placed in an "Agent" folder within Grafana. (@cyriltovena)
-
The Prometheus remote write protocol will now send scraped metadata (metric name, help, type and unit). This results in almost negligent bytes sent increase as metadata is only sent every minute. It is on by default. (@gotjosh)
These metrics are available to monitor metadata being sent:
prometheus_remote_storage_succeeded_metadata_total
prometheus_remote_storage_failed_metadata_total
prometheus_remote_storage_retried_metadata_total
prometheus_remote_storage_sent_batch_duration_seconds
andprometheus_remote_storage_sent_bytes_total
have a new label “type” with the values ofmetadata
orsamples
.
- The Agent has upgraded its vendored Prometheus to v2.17.1 (@rfratto)
-
Invalid configs passed to the agent will now stop the process after they are logged as invalid; previously the Agent process would continue. (@rfratto)
-
Enabling host_filter will now allow metrics from node role Kubernetes service discovery to be scraped properly (e.g., cAdvisor, Kubelet). (@rfratto)
-
Nits in documentation (@sh0rez)
-
Fix various dashboard mixin problems from v0.1.0 (@rfratto)
-
Pass through release tag to
docker build
(@rfratto)
First release!
- Support for scraping Prometheus metrics and sharding the agent through the
presence of a
host_filter
flag within the Agent configuration file.