Configuration reload in 0.2.0
:::warning Diff-classified graceful reload is unavailable
The no-op, dynamic, drain-and-reconcile, and recreate-required reload classifier was added after 0.2.0. The pinned 0.2.0 runtime does not provide that behavior.
:::
Treat a 0.2.0 connection or pool configuration change as a deployment change that can replace pods and interrupt client or upstream connections.
Before changing configuration
- Record the active graph revision and current release.
- Check client, upstream, and idle connection counts.
- Confirm application retry and timeout behavior.
- Validate the change in an isolated environment.
- Keep the previous configuration and release available.
Apply and verify
Use the supported Manager and console workflow. Observe pod replacement, readiness, connection errors, upstream capacity, and application latency. Confirm the active graph revision after reconciliation.
Do not assume that an idle-only drain or in-place ticker update will occur. Those are later-release capabilities.
Rollback
Restore the previous approved configuration and, if required, the previous supported runtime release. Verify client reconnection, MongoDB health, and the active graph before ending the incident or maintenance window.