Helm upgrade for opensearch chart does not disable shard reallocation before starting upgrade

[swapnilsvaidya]

Describe the bug While upgrading OpenSearch version using helm upgrade command, it is observed that shard allocation is not disabled before starting upgrade and enabled again after completing upgrade.

To Reproduce Deploy any version of OpenSearch using helm chart and upgrade the same using helm upgrade command.

Expected behavior As mentioned in opensearch docs https://opensearch.org/docs/1.2/upgrade-to/upgrade-to/ shard allocation should be disabled before upgrade and enabled after upgrade.

Chart Name opensearch

Screenshots If applicable, add screenshots to help explain your problem.

Host/Environment (please complete the following information):

• Helm Version: [e.g. 3.x.x]
• Kubernetes Version: [e.g. 1.16.x]

Additional context Add any other context about the problem here.

[peterzhuamazon]

Hi @swapnilsvaidya do you have any workaround or perma fix to this issue per your experience? Thanks.

cc: @DandyDeveloper @TheAlgo

[TheAlgo]

I am not so sure on how it can be feature w.r.t the Chart. We can have some sidecar scripts embedded though to handle it.

[swapnilsvaidya]

Hi @swapnilsvaidya do you have any workaround or perma fix to this issue per your experience? Thanks.

cc: @DandyDeveloper @TheAlgo

only workaround I can see is to trigger disabling shard re allocation manually before triggering rolling upgrade

[sastorsl]

I think this should not be part of the chart as there are quite a few assumptions which must be made (do we have a cluster, do we have indices) and checks to perform (was shard allocation indeed disabled).

Upgrading OpenSearch (as was/is with the original upstream) requires some planning and forethought.