/ping
Use the lightweight pong response for process liveness. It does not prove every dependency or product journey is healthy.
Operations guide
Put evidence and bounds around the running server.
Persist every required data path, size pools and requests, monitor the process and database, rehearse recovery, and introduce clustering only for a tested need.
- /ping
- /statistics
- Resource bounds
Monitor the correct endpoint.
/statistics
Inspect CPU, memory, load, disk, host, web counters, and database pool statistics.
/health
This is the browser-facing administrative application, not the programmatic liveness probe.
Logs and record history
Collect runtime logs centrally and use record-level history for application investigation, without calling it tamper-proof audit.
Bound shared resources.
- Maximum open and idle database connections fit database capacity and expected concurrency.
- Daptin concurrent-request and IP-plus-exact-path rate limits are exercised; request-body, header, and connection limits are enforced at the trusted reverse proxy.
- Cache headers and invalidation are tested on changing resources and private assets.
- Database, local storage, Yjs files, certificates, and configuration survive replacement of a container or process.
- Backup restoration is rehearsed, not inferred from backup completion.
- The latest artifact is tested, its exact binary or image digest is recorded, and schema changes use representative data before promotion.
- Olric discovery, cache, PubSub, and shared counters are failure-tested before multi-instance rollout.
Match operational claims to the deployed release.
Use /ping and runtime statistics as documented
operational signals in the latest release. Check the latest
release notes before relying on behavior visible only on the
source branch. Repository deployment examples are starting
references, not production guarantees.
Choose a latest-release native artifact or the latest image, a durable SQLite/MySQL/PostgreSQL arrangement, intentional TLS termination, and storage appropriate to the application.
Recovery rehearsal
Operate from evidence, not a green process alone.
-
Use
/pingfor process liveness and define deployment readiness around the dependencies the product requires. - Load the database pool until it waits, then confirm limits and alerts behave as intended.
- Restore a backup into a disposable environment and verify records, file storage, and configuration together.
- Test Daptin concurrency and exact-path rate controls with representative traffic. Enforce header, body, and connection timeouts at a trusted reverse proxy for the latest release.
- Rehearse an upgrade and rollback using the latest release before changing production.