aws_network.aws.txt

                                  ┏━━━━━━━━━━━━━━━━━┓
                                  ┃   AWS_NETWORK   ┃
                                  ┗━━━━━━━━━━━━━━━━━┛

SEE ALSO ==>                      #Doc for AWS general (and related)

                                  ┌─────────────────┐
                                  │   IDEMPOTENCY   │
                                  └─────────────────┘

REQ.ClientToken                   #STR. Do not perform if already performed with same STR.
                                  #Meant for idempotence.
                                  #Only for some SERVICEs
                                  #Sometimes called ClientRequestToken
                                  #Only valid for a specific amount of time, e.g. 36h
                                  #Usually limited to specific chars, usually always at least [:alnum:]-
                                  #Often max 128 chars

REQ|RES.RevisionId                #'REVISION_ID'. Changes each time a MACHINE is updated
                                  #Update call fails if not matching
                                  #I.e. can fetch it, then make update call, which fails if there was a concurrent update in-between
                                  #Only for some SERVICEs, e.g. Lambda or Step Functions VERSION

DELETING RESOURCES ==>            #Must delete other resources that use it first

                                  ┌──────────────────────────┐
                                  │   EVENTUAL CONSISTENCY   │
                                  └──────────────────────────┘

EVENTUAL CONSISTENCY ==>          #Most endpoints use eventual consistency:
                                  #  - action happens async, after response
                                  #  - effects takes time to propagate, i.e. different clients might return different results
                                  #After change:
                                  #  - should test existence
                                  #  - with exponential retry delays up to a limit (e.g. 5 mins)
                                  #  - can be slow, i.e. avoid in critical paths

WAIT ==>                          #Most clients (but not REST API) implement wait methods.
                                  #It calls an ACTION repeatedly until a specific condition fulfills.
                                  #Does it NUM times (SERVICE-specific), every NUM seconds (SERVICE-specific)
                                  #Using an exponential backoff, multiplying by 2 each time.
                                  #Documented with the ACTION it repeats
                                  #Client parameters:
                                  #  - same as ACTION
                                  #  - client-specific backoff|cancellation

                                  ┌──────────────────────────────┐
                                  │   EVENTUAL CONSISTENCY SDK   │
                                  └──────────────────────────────┘

waitUntilWAITComplete
 (WAIT_OPTS, COMMAND_OPTS)        #Wait until WAIT complete
 ->>OBJ                           #COMMAND_OPTS are of the underlying COMMAND
OBJ.state                         #STR among:
                                  #  - 'SUCCESS'
                                  #  - 'FAILURE': condition failed
                                  #  - 'TIMEOUT': WAIT_OPTS.maxWaitTime
                                  #  - 'ABORTED': WAIT_OPTS.abortSignal
                                  #  - 'RETRY': too many attempts (NUM is COMMAND-specific)
OBJ.reason                        #STR

WAIT_OPTS.client                  #CLIENT

WAIT_OPTS.minDelay                #NUM (in secs, def: 30) start of exponential backoff
WAIT_OPTS.maxDelay                #NUM (in secs, def: 120) end of exponential backoff
WAIT_OPTS.maxWaitTime             #Max NUM secs total (def: none)

WAIT_OPTS.abortSignal             #ABORT_SIGNAL

                                  ┌──────────────────────────────┐
                                  │   EVENTUAL CONSISTENCY CLI   │
                                  └──────────────────────────────┘

aws SERVICE WAIT-complete
aws SERVICE wait WAIT             #Wait until WAIT complete

                                  ┌───────────┐
                                  │   RETRY   │
                                  └───────────┘

SERVER ERROR RETRIES ==>          #Should retry on 5**, with exponential backoff
                                  #Could be due to throttling, transient network error, outage, etc.
                                  #Different from retries which wait for eventual consistency

                                  ┌──────────────────┐
                                  │   RETRY CONFIG   │
                                  └──────────────────┘

CONFIG.retry_mode                 #One of:
ENVVAR AWS_RETRY_MODE             #  - 'legacy'
                                  #      - retry on socket errors and throttling
                                  #      - exponential backoff multiplying by 2
                                  #  - 'standard' (def): same but also:
                                  #      - retry on quotas and timeouts
                                  #      - exponential backoff max 20s
                                  #  - 'adaptive': same but also:
                                  #      - backoff is error-specific

CONFIG.max_attempts
ENVVAR AWS_MAX_ATTEMPTS           #NUM of retries (def: 5 with retry_mode 'legacy', 3 otherwise)

                                  ┌───────────────┐
                                  │   RETRY SDK   │
                                  └───────────────┘

COPTS.retryMode|maxAttempts       #Like CONFIG.*

amz-sdk-request [C]               #'attempt=NUM; max=NUM' of retries
                                  #Optional. Set by multiple SDKs

amz-sdk-invocation-id [C]         #'UUID' not changed when doing retries
                                  #Optional. Set by multiple SDKs

METADATA.attempts                 #NUM of retries
METADATA.totalRetryDelay          #NUM of retry delay

                                  ┌───────────────┐
                                  │   RETRY CLI   │
                                  └───────────────┘

--debug                           #Prints retries

                                  ┌───────────────┐
                                  │   ABORT SDK   │
                                  └───────────────┘

SOPTS.abortSignal                 #ABORT_SIGNAL, stopping a single ACTION's underlying HTTP request

                                  ┌────────────────────┐
                                  │   TIMEOUT CONFIG   │
                                  └────────────────────┘

CONNECTION TIMEOUT ==>            #Timeout on TCP|TLS connection start
                                  #Depends on CONFIG.defaults_mode:
                                  #  - 'in-region': 1s
                                  #  - else: 3s
                                  #  - 'mobile': 30s

                                  ┌─────────────────┐
                                  │   TIMEOUT CLI   │
                                  └─────────────────┘

--cli-connect-timeout             #NUM (secs, def: 60). Timeout for connection
--cli-read-timeout                #NUM (secs, def: 60). Timeout, excluding connection

                                  ┌─────────────────┐
                                  │   TIMEOUT SDK   │
                                  └─────────────────┘

*HOPTS|SOPTS.requestTimeout       #NUM (in ms, def: 0) to timeout HTTP connection
                                  #See AWS JavaScript doc for *HOPTS|SOPTS

                                  ┌────────────────────────┐
                                  │   PERFORMANCE CONFIG   │
                                  └────────────────────────┘

CONFIG.tcp_keepalive              #BOOL (def: false)

CONFIG.parameter_validation       #BOOL (def: true). Validate request parameters
                                  #Can disable for performance.

                                  ┌─────────────────────┐
                                  │   PERFORMANCE SDK   │
                                  └─────────────────────┘

ENVVAR AWS_NODEJS_CONNECTION
 _REUSE_ENABLED                   #If '1' (def), use a pool of TCP connections

COPTS.requestHandler              #REQ_HANDLER. How HTTP requests are done.
                                  #Def:
                                  #  - Node.js: node-http-handler
                                  #  - browsers: fetch-http-handler
                                  #Can also directly pass *HOPTS instead of REQ_HANDLER:
                                  #  - most SERVICEs use NHOPTS in Node, FHOPTS in browser
                                  #  - some SERVICEs use N2HOPTS in Node, FHOPTS in browser
                                  #     - Kinesis, Lex Runtime v2, QBusiness, TranscribeStreaming
                                  #Cached, i.e. if multiple CLIENTs, should:
                                  #  - pass same REQ_HANDLER instance as COPTS.requestHandler
                                  #  - not use *HOPTS shortcut syntax

@aws-sdk/node-http-handler        #Use Node HTTP
new NodeHttpHandler([NHOPTS])     #REQ_HANDLER
NHOPTS.http[s]Agent               #AGENT|AGENT_OPTS
                                  #E.g. can set AGENT_OPTS.maxSockets NUM to parallelize HTTP requests

@aws-sdk/node-http2-handler       #Use Node HTTP2
new NodeHttp2Handler
 ([N2HOPTS])                      #REQ_HANDLER
N2HOPTS.sessionTimeout            #Node HTTP2_SESSION.setTimeout()
N2HOPTS.maxConcurrentStreams      #Like Node HTTP2 maxConcurrentStreams
N2HOPTS
 .disableConcurrentStreams        #BOOL (def: false)

@aws-sdk/fetch-http-handler       #Use fetch()
new FetchHttpHandler([FHOPTS])    #REQ_HANDLER

@aws-sdk/xhr-http-handler         #Use XHR
new XhrHttpHandler([XHOPTS])      #REQ_HANDLER

                                  ┌────────────────┐
                                  │   TLS CONFIG   │
                                  └────────────────┘

CONFIG.ca_bundle
ENVVAR AWS_CA_BUNDLE              #'PATH.pem'. Override CA certificate bundle

                                  ┌─────────────┐
                                  │   TLS CLI   │
                                  └─────────────┘

--no-verify-ssl                   #BOOL (def: false)

--ca-bundle                       #Like CONFIG.ca_bundle

--debug                           #Prints TLS info

                                  ┌─────────────────┐
                                  │   TLS LOGGING   │
                                  └─────────────────┘

LOG_EVENT.tlsDetails              #TLS_DETAILS. With CloudTrail
TLS_DETAILS.tlsVersion            #'TLSv1.*'
TLS_DETAILS.cipherSuite           #STR, e.g. 'ECDHE-RSA-AES128-GCM-SHA256'
TLS_DETAILS
 .clientProvidedHostHeader        #Host [C], e.g. 'SERVICE_DOMAIN'

MANAGED_TRAIL_DASHBOARD           #Includes QUERYs:
 Management_Events                #  - TLS versions of last requests
                                  #  - top IDENTITYs using TLS <1.2

                                  ┌────────────────────┐
                                  │   USER AGENT API   │
                                  └────────────────────┘

x-amz-user-agent [C]              #Optional. Set by multiple SDKs
                                  #Includes multiple info for analytics:
                                  #  - SDK, e.g. 'aws-sdk-js X.Y.Z'
                                  #  - runtime, e.g. 'md/nodejs X.Y.Z'
                                  #  - OS, e.g. 'os/linux X.Y.Z-A-generic'
                                  #  - APPLICATION_ID (see AWS identifiers doc)
                                  #  - opt-in features (as letters), e.g. S3 express, ACCOUNT-based endpoint, etc.

                                  ┌────────────────────┐
                                  │   USER AGENT SDK   │
                                  └────────────────────┘

COPTS.customUserAgent             #STR|USER_AGENTS. x-amz-user-agent [C]
USER_AGENTS                       #['NAME'[, VERSION]]_ARR

                                  ┌────────────────────┐
                                  │   USER AGENT CLI   │
                                  └────────────────────┘

--debug                           #Prints user agent

                                  ┌────────────────────────┐
                                  │   USER AGENT LOGGING   │
                                  └────────────────────────┘

LOG_EVENT.userAgent               #STR of User-Agent [C]. With CloudTrail
                                  #Can also be:
                                  #  - 'signin.amazonaws.com': UI console, as USER
                                  #  - 'console.amazonaws.com': UI console, as root
                                  #  - 'lambda.amazonaws.com': inside a Lambda FUNCTION
                                  #  - 'AWS Internal/INTERNAL_ID'

                                  ┌────────────┐
                                  │   QUOTAS   │
                                  └────────────┘

QUOTAS ==>                        #Also named "limits"
                                  #"Flexible": when can increase through support request
                                  #Usually per REGION

                                  ┌────────────────┐
                                  │   THROTTLING   │
                                  └────────────────┘

THROTTLING ==>                    #Throttling quota is ACTION-specific, sometimes also REGION-specific
                                  #Error 'CODE' is also ACTION-specific

                                  ┌────────────────────┐
                                  │   THROTTLING SDK   │
                                  └────────────────────┘

AWS_ERROR.$retryable.throttling   #BOOL. Was throttled

                                  ┌────────────────────────┐
                                  │   THROTTLING LOGGING   │
                                  └────────────────────────┘

MANAGED_TRAIL_DASHBOARD
 Highlights                       #Includes QUERY: throttled requests with management EVENTs
MANAGED_TRAIL_DASHBOARD
 Overview|Error_Analysis_Dashboard#Includes QUERY: top throttled ACTIONs
MANAGED_TRAIL_DASHBOARD
 DataEvents_Overview_Dashboard    #Includes QUERY: top throttled ACTIONs with data EVENTs