Upgrading Slurm Bridge
Supported upgrade paths
A release series is identified by the first two version numbers, such as 1.2.
Supported upgrade paths within the same first version number are:
Upgrade |
Supported path |
|---|---|
|
Upgrade directly to any newer patch; intermediate patches may be skipped. |
|
Upgrade directly to the next series. |
|
Upgrade through |
Use the latest release in each series. Complete and verify each step before proceeding to the next release series.
Before upgrading
Follow the upgrade policy: pause submissions and drain pending Bridge work before starting the upgrade. The Helm hook checks pending Pods before stopping Bridge components; Slurm queue verification remains an operator step.
Check the installed release’s chart version and application version:
helm get metadata slurm-bridge --namespace slurm --output json
Compare these versions with the intended target release. Helm’s revision number counts release operations; it is not the Slurm Bridge version. Keep the chart and all three bridge component images on the intended application release.
Review the target release’s configuration changes and confirm that your Kubernetes and Slurm versions are compatible with both releases. After each upgrade, wait for all bridge components to become ready and verify that workers are registered with Slurm and a test workload can run and release its allocation.
The Helm chart and running components do not currently reject skipped release series. A successful Helm upgrade or healthy pod does not establish that an upgrade path is supported. Check the source and target versions before applying the upgrade.
For release validation on a dedicated cluster, the TestUpgrade e2e spec checks
the version path before changing the release, preserves running workloads across
the upgrade, rejects an upgrade with pending work, and verifies that new work
can run after upgrading with a drained queue.