Skip to content

All index info lost when the original master machines are not participating in master election #9975

Closed
@mkliu

Description

@mkliu

It appears to be as resiliency problem, and i haven't seen it mentioned any place obvious so hopeful it could get fixed.

We had a problem recently about master nodes failing, so I enabled my data nodes to serve as master too - dualnode. However, even though the cluster name is the same, i don't see any of my previous indexes anymore! After quite a while I realize probably because the global cluster status is saved on master node's disk, so the new dualnode doesn't know. I switched back to the old masters when they come back up, then indexes appeared again, but many of them are in red state....

My question would be, since the index and segments are still on the data node, why can't the dualnode detect the indexes that's already in the cluster?

This is really a nasty trap...

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions