Upgrade to a later SDP release

Upgrade the Stackable Data Platform (SDP) from the release you run to a later one. The upgrade planner on the Stackable Hub can be used to generate a detailed plan for your specific version combination.

Two rules from the upgrade policy are important to keep in mind for _every update:

Skipping SDP releases is not supported. Upgrade from one SDP release to the next, for example 24.11 → 25.3 → 25.7.
Upgrade the operators first and the product versions afterwards. Every product version is deprecated for at least one SDP release before it is removed, so the operators can move to the next release while the products stay on the versions they run.

If you use custom images, the cluster definitions have to change together with the operators, see custom images.

Upgrade to the next release

Repeat these steps for every release between the one you run and the one you want. The commands for each release and any steps it needs by hand are in the relevant upgrade sections of the release notes.

Read the release notes of the release you upgrade to, including the upgrade section and the product upgrade guides it links. Some of these steps have to happen before the operators are replaced.
Move every stacklet whose product version the next release no longer ships to a version both releases support, by changing its spec.image.productVersion. Do the same for the Kubernetes or OpenShift version of the cluster. The release notes and the Stackable Hub list the product and platform versions each release supports.
Pause reconciliation of every stacklet, see Cluster operations:
```
$ kubectl patch <kind>/<name> --type=merge --patch '{"spec": {"clusterOperation": {"reconciliationPaused": true}}}'
```
Upgraded operators automatically and immediately restarts the stacklets they manage using the images of the new release, see Product image selection. Pausing reconciliation lets you restart them one at a time at your own pace instead of all at once.
Upgrade the operators as the release notes describes, for example with stackablectl release upgrade <release>. Replace the CustomResourceDefinitions (CRDs) by hand if the release notes say so.
Resume reconciliation of one stacklet:
```
$ kubectl patch <kind>/<name> --type=merge --patch '{"spec": {"clusterOperation": {"reconciliationPaused": false}}}'
```
The operator restarts the stacklet using the new images. Wait until all its Pods are ready again and the operator logs show no errors, then resume the next one.
Change spec.image.productVersion where you want a newer product version. Each change restarts that stacklet once more. Some products need a manual process for a version change, such as Apache Hadoop HDFS.