[FEATURE] Disable ML SyncUp by default when running cluster on single-node mode

[eusousu]

Is your feature request related to a problem?
As discussed here on opensearch forum, when ml is enabled there is a periodic SyncUp job that executes to maintain your ml model status across the cluster in data nodes. By default running every 10 seconds. This seem unnecessary when the cluster is running on single node mode or when there is only one ML node.

What solution would you like?
There is an option to disable this SyncUp by setting plugins.ml_commons.sync_up_job_interval_in_seconds to 0, this could be the default when running on single-node mode or when there is only one ML node.

Do you have any additional context?
Relevant discussion on forum: https://forum.opensearch.org/t/high-rate-of-queries-to-plugins-ml-model-is-this-normal/24845

[pyek-bot]

Hi @eusousu , is this amazon managed opensearch?

[Zhangxunmt]

Hi @eusousu , this syncUp job is still necessary for single node cluster to update the model status. In the case when the single node is dropped, the syncUp job updates the model status to "failed" so you would know that the model is not available anymore. Do you have any specific concern for having a job running in your cluster? Just treat this as another job stated through the job-scheduler plugin.