merge col-exporter-rewrite into main (#295)

* Skip all fixture tests (#239) * Initial structure for new pdata metrics exporter (#238) * [Metrics Rewrite] add outline with todos for fragmenting work (#240) * [Metrics Rewrite] attribute to label mapping (#243) [Metrics Rewrite] attribute to label mapping * [Metrics Rewrite] support for pdata Sum points (#242) * [Metrics Rewrite] support for pdata Sum points * update breaking-changes.md * use concatentation instead of sprintf * [Metrics Rewrite] support for pdata Gauge points (#244) * Add logic to translate metric descriptors and initial flow (#247) * Fixes from merge. * Fix tests. * Clean up test cases, re-disable integration tests. * Add summary descriptors and label descriptors. * Fix lint issues. * Some fixes from review. * Remove metric import. * Fixes from review. - Update default config method - Simplify some of my lack-of-go expertise. * Add unit test for metric domains. * Fixes from review. * Add breaking changes. * Fixes from review. * Update context to be TODO. * Add support for exponential histograms and exemplars. (#251) * Add support for exponential histograms and exemplars. * Fixes from review. * Fixes from review. * Fixes from discussion. * [Metrics Rewrite] implement monitored resource mapping (#252) * [Metrics Rewrite] implement monitored resource mapping * review fixes * [Metrics Rewrite] update breaking-changes.md for monitored resource (#255) * Add summary mapping to exporter. (#249) * Add config to call `CreateServiceTimeSeries` (#259) * Initial implementation of create service time series. * Add a test case for create service timeseries. * Add logic to auto-detect project id if not configured. * Fix from code review * Fix resource to be one that has retention policy for integration tests. * Add support for histogram to metrics exporter. (#258) BUG=210164184 * Re-enable ops-agent self-metric integration test. (#260) * [Metrics Rewrite] add ExponentialHistogram fixture (#257) * [Metrics Rewrite] add ExponentialHistogram fixture * make tests deterministic * few last changes * close channel instead of sending a message * Enable ops agent host metric integration test. (#264) - There is a bug in upstream agent-metric-processor that sets incorrect units on usage metrics (GoogleCloudPlatform/opentelemetry-operations-collector#72) - We update the expectations for inculsion of units in CreateTimeSeries - We disable metric descriptors (for now). Given the bug in agent-metric-processor, liekly ops-agent will need upstream fix for this first. * add a feature gate, which defaults to false, for using the re-written exporter (#267) * Enable Basic integration tests (#266) * Enable basic counter test. * Enable delta counter metrics. - Note: Delta counters are NOW fake-delta (i.e. cumulatives with limited time windows) * Enable non-monotonic-sum integration test. * Re-enable summary integration test and fix design issues in summary translation. - Summary exports percentiles, not quantiles - Percentiles should include similar double precision in the string. * Fix recordfixtures script to use featuregate (#270) * Skip already seen attribute keys when creating LabelDescriptors (#272) * Reenable GKE metrics agent fixtures (#271) * Update breaking-changes.md for googlecloudmonitoring/point_count self observability (#277) * Move logging to use zap-logger and set up self-observability to match collector expectations. (#275) * Enable metric prefix integraiton tests. (#274) * enable workloadapis prefix integration test. * update unknown domain metrics expect. * Add instrumentationLibraryToLabels method to metrics exporter. (#253) * Add instrumentationLibraryToLabels method to metrics exporter. BUG=https://b.corp.google.com/issues/210164355 * Remove custom_metrics_domains behaviour from metrics-exporter. * Remove dependency on go.opentelemetry.io/collector (#279) * remove dependency on go.opentelemetry.io/collector * add ocgrpc metrics to exporters' self-obs metrics (#280) * Use OC stackdriver exporter to capture self observability metrics as GCM protos (#282) * Capture ocgrpc self observability metrics (#283) * make integrationtest not internal (#285) * Remove internal/ prefix for integrationtest (#288) * Add batching support to metrics-exporter. (#286) * Add batching support to metrics-exporter. * Retry when we fail to write metric descriptors. * Re-enable workload metrics integration tests (#278) * update header year for new files (#296) * Document new CreateMetricDescriptor behavior (#294) * reenable disabled metrics test (#299) Co-authored-by: Aaron Abbott <[email protected]> Co-authored-by: Josh Suereth <[email protected]> Co-authored-by: Thomas Barker <[email protected]> Co-authored-by: Punya Biswal <[email protected]>
GoogleCloudPlatform · Feb 2, 2022 · 7338096 · 7338096
1 parent 242ff20
commit 7338096
Show file tree

Hide file tree

Showing 67 changed files with 22,825 additions and 2,076 deletions.
diff --git a/Makefile b/Makefile
@@ -16,7 +16,7 @@ ifeq ($(UNAME_S),Darwin)
 	endif
 endif
 
-GOTEST_MIN = go test -v -timeout 60s
+GOTEST_MIN = go test -v -timeout 70s
 GOTEST = $(GOTEST_MIN) -race
 GOTEST_WITH_COVERAGE = $(GOTEST) -coverprofile=coverage.txt -covermode=atomic
 

diff --git a/exporter/collector/README.md b/exporter/collector/README.md
@@ -201,7 +201,6 @@ and [memory limiter](https://github.com/open-telemetry/opentelemetry-collector/t
 optimal network usage and avoiding memory overruns.  You may also want to run an additional
 [sampler](https://github.com/open-telemetry/opentelemetry-collector/tree/main/processor/probabilisticsamplerprocessor), depending on your needs.
 
-
 ## Deprecatations
 
 The previous trace configuration (v0.21.0) has been deprecated in favor of the common configuration options available in OpenTelemetry. These will cause a failure to start

diff --git a/exporter/collector/breaking-changes.md b/exporter/collector/breaking-changes.md
@@ -0,0 +1,88 @@
+# Breaking changes vs old googlecloud exporter
+
+The new pdata based exporter has some breaking changes from the original OpenCensus (OC)
+stackdriver based `googlecloud` exporter:
+
+## Metric Names and Descriptors
+
+The previous collector exporter would default to sending metrics with the type:
+`custom.googleapis.com/OpenCensus/{metric_name}`.  This has been changed to
+`workload.googleapis.com/{metric_name}`.
+
+Additionally, the previous exporter had a hardcoded list of known metric domains
+where this "prefix" would not be used. The new exporter allows full configuration
+of this list via the `metric.known_domains` property.
+
+The previous exporter would, by default, only call CreateMetricDescriptor for metrics with
+domain `custom.googleapis.com` or `external.googleapis.com`. The new exporter will try to call
+CreateMetricDescriptor regardless of domain, unless `metric.skip_create_descriptor` or
+`metric.create_service_timeseries` are true. The exporter now calls CreateMetricDescriptor as a
+best effort. It queues MetricDescriptors in a buffered channel, dropping them when the channel
+is full; the next time that metric is seen, it will be retried.
+
+Additionally, the DisplayName for a metric used to be exactly the
+`{metric_name}`. Now, the metric name is chosen as the full-path after the
+domain name of the metric type.  E.g. if a metric called
+`workload.googleapis.com/nginx/latency` is created, the display name will
+be `nginx/latency` instead of `workload.googleapis.com/nginx/latency`.
+
+## Monitored Resources
+
+Mapping from OTel Resource to GCM monitored resource has been completely changed. The OC based
+exporter worked by converting the OTel resource into an OC resource which the exporter
+recognized. The `resource_mappings` config option allowed customizing this conversion so the OC
+exporter would correctly convert to a GCM monitored resource.
+
+Then new pdata based exporter works by interpreting the [OTel Resource semantic
+conventions](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/resource/semantic_conventions/README.md)
+as follows to determine the monitored resource type:
+
+- Switch on the `cloud.platform` Resource attribute and if:
+  - `gcp_compute_engine`, send a `gce_instance` monitored resource.
+  - `gcp_kubernetes_engine`, send the most specific possible k8s monitored resource depending
+  on which resource keys are present and non-empty. In order, try for `k8s_container`,
+  `k8s_pod`, `k8s_node`, `k8s_cluster`.
+  - `aws_ec2`, send a `aws_ec2_instance` monitored resource.
+- Otherwise, fallback to:
+  - `generic_task` if the `service.name` and `service.instance_id` resource attributes are
+  present and non-empty.
+  - `generic_node`
+
+Once the type is determine, the monitored resource labels are populated from the mappings
+defined in [`monitoredresource.go`](monitoredresource.go#L51). The new behavior will never send the
+`global` monitored resource.
+
+For now, it is not possible to customizate the mapping algorithm, beyond using the
+[`resourceprocessor`](https://github.com/open-telemetry/opentelemetry-collector-contrib/tree/main/processor/resourceprocessor)
+in the collector pipeline before the exporter. If you have a use case for customizing the
+behavior, please open an issue.
+
+## Labels
+
+Original label key mapping code is
+[here](https://github.com/census-ecosystem/opencensus-go-exporter-stackdriver/blob/42e7e58efdb937e8477f827d3fba022212335dbc/sanitize.go#L26).
+The new code does not:
+
+- truncate label keys longer than 100 characters.
+- prepend `key` when the first character is `_`.
+
+## OTLP Sum
+
+In the old exporter, delta sums were converted into GAUGE points ([see test
+fixture](https://github.com/GoogleCloudPlatform/opentelemetry-operations-go/blob/9bc1f49ebe000b0b3b1aa5b7f201e7996effdcd8/exporter/collector/testdata/fixtures/delta_counter_metrics_expect.json#L15)).
+The new pdata exporter sends these as CUMULATIVE points with the same delta time window
+(reseting at each point) aka pseudo-cumulatives.
+
+## OTLP Summary
+
+The old exporter relied on upstream conversion of OTLP Summary into Gauge and
+Cumulative points.  The new exporter performs this conversion itself, which
+means summary metric descriptors will include label description for `percentile`
+labels.
+
+## Self Observability Metrics
+
+For each OTLP Summary metric point, the old exporter would add 1 to the
+`googlecloudmonitoring/point_count` self-observability counter. For a Summary point with N
+percentile values, the new exporter will add `N + 2` (one for each percentile timeseries, one
+for count, and one for sum) to the counter.
diff --git a/exporter/collector/config.go b/exporter/collector/config.go
@@ -15,25 +15,23 @@
 package collector
 
 import (
-	"go.opentelemetry.io/collector/config"
-	"go.opentelemetry.io/collector/exporter/exporterhelper"
+	"time"
+
 	"google.golang.org/api/option"
 )
 
+const (
+	DefaultTimeout = 12 * time.Second // Consistent with Cloud Monitoring's timeout
+)
+
 // Config defines configuration for Google Cloud exporter.
 type Config struct {
-	config.ExporterSettings `mapstructure:",squash"`
-	ProjectID               string `mapstructure:"project"`
-	UserAgent               string `mapstructure:"user_agent"`
-	Endpoint                string `mapstructure:"endpoint"`
+	ProjectID string `mapstructure:"project"`
+	UserAgent string `mapstructure:"user_agent"`
+	Endpoint  string `mapstructure:"endpoint"`
 	// Only has effect if Endpoint is not ""
 	UseInsecure bool `mapstructure:"use_insecure"`
 
-	// Timeout for all API calls. If not set, defaults to 12 seconds.
-	exporterhelper.TimeoutSettings `mapstructure:",squash"` // squash ensures fields are correctly decoded in embedded struct.
-	exporterhelper.QueueSettings   `mapstructure:"sending_queue"`
-	exporterhelper.RetrySettings   `mapstructure:"retry_on_failure"`
-
 	ResourceMappings []ResourceMapping `mapstructure:"resource_mappings"`
 	// GetClientOptions returns additional options to be passed
 	// to the underlying Google Cloud API client.
@@ -47,6 +45,20 @@ type Config struct {
 type MetricConfig struct {
 	Prefix                     string `mapstructure:"prefix"`
 	SkipCreateMetricDescriptor bool   `mapstructure:"skip_create_descriptor"`
+	// If a metric belongs to one of these domains it does not get a prefix.
+	KnownDomains []string `mapstructure:"known_domains"`
+
+	// If true, set the instrumentation_source and instrumentation_version
+	// labels. Defaults to true.
+	InstrumentationLibraryLabels bool `mapstructure:"instrumentation_library_labels"`
+
+	// If true, this will send all timeseries using `CreateServiceTimeSeries`.
+	// Implicitly, this sets `SkipMetricDescriptor` to true.
+	CreateServiceTimeSeries bool `mapstructure:"create_service_timeseries"`
+
+	// Buffer size for the channel which asynchronously calls CreateMetricDescriptor. Default
+	// is 10.
+	CreateMetricDescriptorBufferSize int `mapstructure:"create_metric_descriptor_buffer_size"`
 }
 
 // ResourceMapping defines mapping of resources from source (OpenCensus) to target (Google Cloud).
@@ -64,3 +76,19 @@ type LabelMapping struct {
 	// When required label is missing, we fallback to default resource mapping.
 	Optional bool `mapstructure:"optional"`
 }
+
+// Known metric domains. Note: This is now configurable for advanced usages.
+var domains = []string{"googleapis.com", "kubernetes.io", "istio.io", "knative.dev"}
+
+// DefaultConfig creates the default configuration for exporter.
+func DefaultConfig() Config {
+	return Config{
+		UserAgent: "opentelemetry-collector-contrib {{version}}",
+		MetricConfig: MetricConfig{
+			KnownDomains:                     domains,
+			Prefix:                           "workload.googleapis.com",
+			CreateMetricDescriptorBufferSize: 10,
+			InstrumentationLibraryLabels:     true,
+		},
+	}
+}
diff --git a/exporter/collector/config_test.go b/exporter/collector/config_test.go
diff --git a/exporter/collector/factory.go b/exporter/collector/factory.go