HDDS-4896. Need a tool to upgrade current non-HA SCM node to single node HA cluster

[bshashikant]

What changes were proposed in this pull request?

The idea here is to add an optional config "ozone.scm.primordial.node.id". if the config is set, it will cause scm --init to only take effect on
the specific node and ignore scm --bootstrap cmd. Similarly, scm --init will be ignored on the non-primordial scm nodes.

If the cluster is upgraded from non-ratis to ratis version , scm --init needs to re-run for switching from non-ratis based SCM to ratis-based SCM on the primary node.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-4896

How was this patch tested?

Unit tests added

[GlenGeng-awx]

How about move validateRatisGroupExists to TestStorageContainerManager, since it is not used in production code any more.

[GlenGeng-awx]

One question about the necessity of the optional config ozone.scm.primordial.node.id, In which situation will this flag be used ?

With this flag, people can apply init, bootstrap and start on all of the 3 SCMs, so that manual operation will be safer. Is this the requirement ?

[bshashikant]

One question about the necessity of the optional config ozone.scm.primordial.node.id, In which situation will this flag be used ?

With this flag, people can apply init, bootstrap and start on all of the 3 SCMs, so that manual operation will be safer. Is this the requirement ?

yes. This is a requirement for Kubernetes setups .

[GlenGeng-awx]

One question about the necessity of the optional config ozone.scm.primordial.node.id, In which situation will this flag be used ?
With this flag, people can apply init, bootstrap and start on all of the 3 SCMs, so that manual operation will be safer. Is this the requirement ?

yes. This is a requirement for Kubernetes setups .

Thanks for the explanation. Could you add the explanation for the scenario in comments and ozone-site.xml ?

[GlenGeng-awx]

+1.