Skip to main content

​
Upgrading

Run helm upgrade with the new chart version and your values file:
Workers finish their current unit before they are replaced, the orchestrator resumes its queue, and the pool keeps its size. While the orchestrator restarts, the API answers 503 with Retry-After. With Helm 4, an upgrade can fail with a conflict on .spec.replicas if the pool resized during it. Run the same command again. Do not force it.

​
Changes that need a reinstall

Some storage settings cannot change on an existing release: storage.inProcess.efs.fileSystemId, storage.inProcess.storageClass.*, storage.output.s3.bucket, storage.s3.region, and storage.s3.mountOptions. Let the queue drain, uninstall, and install again with the new values.

​
Uninstalling

Let running jobs finish first, then:
The orchestrator’s state claim, mnesia-fineparser-orchestrator-0, is kept. A new install with the same release name and namespace picks up its queue. To start empty, delete it:
The chart-created EFS access point is deleted, and by default its data stays on the file system. The EFS file system, the S3 bucket and its results, and any IAM role you created are kept.