Drop rules & retention
Stop keeping the logs you don't need, so they don't fill the Logs page or count toward your monthly allowance — and know how long the rest is kept.
Drop rules
Some logs you don't want kept — a cluster's system pods (aws-ebs-csi-driver, cainjector), health-check noise. A drop rule drops matching lines the moment they arrive, from every agent, SDK, log source and API call: they aren't stored, don't appear in the live tail and don't count toward your monthly allowance.
- One service: on the Logs page, open the services menu and choose Exclude beside it, then Exclude service.
- A pattern: Excluded logs… at the bottom of the services menu adds a rule by Service, Environment or Source — the whole name, with
*matching anything, as inaws-ebs-*— or by Message contains, as inGET /healthz. - Keep beside an excluded service brings it back, from then on.
Rules take effect within 30 seconds. Lines already stored stay until they expire. Owners and managers can change the rules; there's also an API.
To stop logs before they leave a server at all, exclude them in the agent instead — a container label, a pod annotation or a Kubernetes namespace: see Log Configuration. In an SDK, lower what it sends — for example a higher logging level.
Retention and allowance
| Plan | Kept for | Sent per month |
|---|---|---|
| Free | 7 days | 3 GB |
| Pro | 30 days | 30 GB |
| Business | 90 days | 200 GB |
- Every log is kept for the plan's retention, whichever way it arrived; logs on an app's Logs tab are the same lines and expire with them.
- The allowance is counted per calendar month (UTC); Enterprise has none. A batch that crosses the line is still accepted; after that, new lines are refused (
429) until the 1st. This month's use is in GET /workspaces/:id/usage. - Dropped lines never count.