[Elasticsearch-OSS] Cannot Start using OSS version

[ichasco-heytrade]

Hi, I am having an error when deploying elasticsearch-oss because is not compatible with node.ml. And the script force to add it.
The only way to run it is to edit manually the statefullset and delete:

containers:
      - env:
        - name: node.ml
          value: false,

It doesn't matter if I set it to false. It crash with this error:

"stacktrace": ["org.elasticsearch.bootstrap.StartupException: java.lang.IllegalArgumentException: unknown setting [node.ml] please check that any required plugins are installed, or check the breaking changes documentation for removed settings",

Chart version: 7.11.1

Kubernetes version: v1.18.9

Kubernetes provider: EKS

Helm Version: v3.5.0

Elasticsearch Version: 7.10.2

Thanks!

[xrl]

I had the same issue and I read over the templates a bit, looks like they take the entries from the release's role list and dynamically create ES config properties. The new default values include a ml and remote_cluster_client whose presence, true or false, would put an unsupported config in to my version of ES (7.2.0 which I'm in the process of upgrading, hence this issue).

The fix? Don't let the helm template see any value for those two bad keys:

# snippet from my logs-master.yaml value file
roles:
  master: "true"
  ingest: "false"
  data: "false"
  ml:
  remote_cluster_client:

[ichasco-heytrade]

Thanks @xrl it seems to work like that! :)

[xrl]

Not positive this should be closed, this is a nasty bug for future users :)

[ichasco-heytrade]

Ok I will let it opened

[jmlrt]

Hi @ichasco-heytrade, @xrl,
Elastic charts provide default values and support when used with the same version for chart and application.
This is why the roles value is expected to work with Elasticsearch 7.11 but not necessarily with previous versions or OSS version which doesn't exist anymore in 7.11.

That's why we won't update the values and your workaround is the good to handle it on your side.

[xrl]

This chart says it supports the 6-series. At the very least it could be documented?

[jmlrt]

Indeed, that's why we have a different branch for 6.8 which doesn't include ml and remote_cluster_client in default values + tags for every releases and we specify here that you should use an Helm chart version is aligned to the version of the product you want to deploy.

We recommend that the Helm chart version is aligned to the version of the product you want to deploy. This will ensure that you using a chart version that has been tested against the corresponding production version. This will also ensure that the documentation and examples for the chart will work with the version of the product you are installing.

For example if you want to deploy an Elasticsearch 7.7.1 cluster, use the corresponding 7.7.1 tag.

[botzill]

I have similar issue trying to install latest version 8-snapshot, I get:

unknown setting [node.master] 

Is there a way to bypass this and still test/use version 8?

Seems that docker image 8.0.0-SNAPSHOT is replaced when a new update is made, as this was working for me some time ago.

[jmlrt]

Hi @botzill,using Elasticsearch 8.0.0-SNAPSHOT with this chart is currently broken because of #855 and we have a wip PR to fix it: #1186

Please note that Elasticsearch 8.0.0-SNAPSHOT is an unsupported development version available only for tests.

The latest supported version is 7.12.1 for now.

[botelastic]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[botelastic]

This issue has been automatically closed because it has not had recent activity since being marked as stale.