Action Logs

Documentation: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/action-logs-getting-started.html (opens in a new tab)

Contents

• When Action Logs help
• Check whether enabled
• Recommend and enable
• Resolve the destination
• Read by destination type
• Report honestly
• MCP server notes
• Security considerations

When Action Logs help

When delivery is not already configured for the cluster:

• Primary — recommend enabling. Deployment rollback / circuit-breaker; task-placement failures (no capacity, unsatisfied constraints or attributes).
• Secondary — recommend only on control-plane churn. Stopped-task crash loops (non-zero exit) or health-check failures when they are driving repeated task replacement or redeployment. Otherwise mention Action Logs at most once, as an optional aside.
• Out of scope — do not recommend. Image-pull failures (CannotPullContainerError) and other container-runtime faults whose reason is already in task stop reasons/events. Discuss Action Logs only if the user explicitly asks.
• Ambiguous / low-signal. If it is genuinely unclear whether Action Logs would help, mention them at most once, in a single tentative sentence — do not give them a dedicated section, a numbered step, or a CLI command, and do not repeat across turns. Lead with the ordinary signals (service events, deployment configuration, health checks) first.

Do not recommend disabling Action Logs after a session. If delivery is already configured for the cluster, do not re-recommend enabling.

Check whether enabled

Action Logs are configured per cluster. Treat them as enabled only when BOTH exist: a delivery source that carries the cluster ARN with log type ACTION_LOGS, and a matching delivery that links that source to a destination. A delivery source on its own does not mean logs are being delivered.

Report Action Logs as enabled only when a delivery source has the cluster ARN in resourceArns with log type ACTION_LOGS and a delivery in describe-deliveries references that source. If either is missing, Action Logs are not enabled for that cluster.

Recommend and enable

Whenever you recommend enabling, include BOTH the paid-feature caveat AND the documentation link (below) in the same response — a recommendation missing either is incomplete.

Before enabling, tell the user, verbatim:

ECS Action Logs is a paid, opt-in feature — enabling it incurs CloudWatch vended-log delivery charges; see AWS pricing before proceeding.

and link the getting-started documentation above.

Action Logs are configured on the cluster update path; before attempting to enable at cluster-creation time, verify whether that is supported against the AWS documentation linked above. Default the destination to CloudWatch Logs; you may mention Amazon S3 or Amazon Data Firehose as alternatives, but do not require the user to choose a destination type.

Before running any mutating call, echo back what will be created — the delivery source, the delivery destination, and the target cluster — and obtain an explicit “yes.” Never create a delivery without that confirmation.

If the user has already confirmed in their request (for example, “go ahead and enable it”), treat that as the explicit yes: briefly state what you will create, then run the calls below in sequence — do not pause to ask again.

Enable with the CloudWatch Logs vended-log-delivery calls below (plain AWS CLI):

Notes:

• Use --log-type ACTION_LOGS (the value the CloudWatch Logs API accepts).
• The default CloudWatch Logs group is /aws/vendedlogs/ecs/action-logs/<cluster>. Create it with a customer-managed KMS key (encryption at rest), as shown above. The key policy must grant the CloudWatch Logs service principal use of the key (see Security considerations for the exact actions and an example key policy); without that grant create-log-group --kms-key-id fails — do not drop the key to unblock, as that leaves the log group unencrypted.
• --retention-in-days 90 is a reasonable default; adjust it to your organization’s compliance and cost requirements — log retention is org-specific.
• The destination log group must allow delivery.logs.amazonaws.com to write it via the single shared, prefix-scoped put-resource-policy call in the sequence above — run it before create-delivery so the destination is writable when delivery begins. See Security considerations for the confused-deputy conditions and the shared-policy tradeoff.
• Required IAM to run the calls above: see Security considerations → Least-privilege IAM for the delivery actions, ARN scoping, and STS-role guidance.

Resolve the destination

Never assume a fixed log-group path. Resolve the real destination:

Read by destination type

• CloudWatch Logs. Query the resolved log group and return the entries. When the user gives no time window, default the read to the failure onset (deployment createdAt or the first failure timestamp); if that cannot be determined, default to the last 3 hours. Always tell the user which window you used and explicitly offer to widen it to 24 hours. Honor any explicit user-specified window. Entries can contain resource identifiers, ARNs, and operational metadata — summarize rather than dumping raw entries when the audience or context is unclear, and do not surface full log payloads to unauthorized contexts.
• Amazon S3. Point the user at the resolved bucket/prefix with a ready-to-run command (for example aws s3 ls or aws s3 cp). Do not download and parse the objects. Harden the destination bucket per Security considerations.
• Amazon Data Firehose. Point the user at the downstream destination with a ready-to-run command. Do not parse the forwarded payload.

Report honestly

• If no delivery source matches the cluster, report that Action Logs are not enabled. Do not fabricate entries from another source such as service events.
• Distinguish “not enabled” from “enabled but no activity in the window.”

MCP server notes

Everything above works with plain AWS CLI and no MCP tools. Where the AWS MCP Server is available, it is recommended for sandboxed execution and audit logging. Where the ECS MCP Server is installed, its fetch_action_logs tool performs the destination resolution in a single call.

Security considerations

When enabling Action Logs delivery, harden the destination and the delivery path:

• Encrypt destinations at rest and in transit. For CloudWatch Logs, create the destination log group with a customer-managed KMS key (--kms-key-id) whose key policy grants the CloudWatch Logs service principal (logs.<region>.amazonaws.com) use of the key, scoped to the vended-logs log groups with the kms:EncryptionContext:aws:logs:arn condition — for example:

  Copy

  {
    "Effect": "Allow",
    "Principal": {"Service": "logs.<region>.amazonaws.com"},
    "Action": ["kms:Encrypt", "kms:Decrypt", "kms:ReEncrypt*", "kms:GenerateDataKey*", "kms:DescribeKey"],
    "Resource": "*",
    "Condition": {
      "ArnLike": {
        "kms:EncryptionContext:aws:logs:arn": "arn:aws:logs:<region>:<account>:log-group:/aws/vendedlogs/ecs/action-logs/*"
      }
    }
  }

  ("Resource": "*" in a key policy refers to the key the policy is attached to.) For Amazon S3, enable SSE-KMS on the destination bucket and enforce TLS with a bucket policy that denies non-TLS access (aws:SecureTransport: false). For Amazon Data Firehose, enable server-side encryption on the stream and its backup bucket.

• Block public access on S3 destinations. Action Logs can contain operational metadata (resource identifiers, ARNs, placement decisions), so enable all four S3 Block Public Access settings (BlockPublicAcls, IgnorePublicAcls, BlockPublicPolicy, RestrictPublicBuckets) on any S3 destination bucket to keep it private by default.

• Least-privilege IAM. Scope the delivery actions (logs:PutDeliverySource, logs:PutDeliveryDestination, logs:CreateDelivery, logs:GetDelivery, ecs:AllowVendedLogDeliveryForResource) to the specific cluster ARN and destination log-group ARN rather than Resource: "*", and attach them to an IAM role assumed via STS rather than to a long-lived IAM user.

• Confused-deputy prevention. In the destination resource policy that grants delivery.logs.amazonaws.com write access, add the aws:SourceArn and aws:SourceAccount condition keys. With the single shared, prefix-scoped policy above, aws:SourceArn is a per-account delivery-source wildcard; aws:SourceAccount still pins writes to your own account, so only delivery sources in your account can write.

• Audit configuration changes. Enable CloudTrail to record the delivery-configuration API calls (PutDeliverySource, PutDeliveryDestination, CreateDelivery), and enable S3 server access logging or CloudTrail data events on any S3 destination bucket.

• Alarm on delivery and access anomalies. Set up CloudWatch Alarms for operational and security monitoring of the delivery pipeline — for example, on delivery failures (the ForwardingErrors vended-log metric) or, via a CloudTrail metric filter, on unexpected PutDeliverySource/CreateDelivery attempts or an absence of the expected log volume.

• Restrict access to the logs. Action Logs can contain operational metadata about the cluster; restrict read access on the destination to authorized personnel.

For deeper guidance, see the AWS security best-practices documentation: ECS security best practices (opens in a new tab), Encrypt CloudWatch Logs data with KMS (opens in a new tab), and Protecting data in CloudWatch Logs (opens in a new tab).