Description
I've found that in this case we'll have both ip and ip_start and ip is latest (because as part of joining cluster somebody trying to assign bad name to it). But when cluster is restarted ip_start takes precedence node starts under old name. Which leads to auto-ejection.
Attachments
For Gerrit Dashboard: MB-8098 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
25667,2 | MB-8098: delete ip_start when ip is updated | master | ns_server | Status: MERGED | +2 | +1 |
26278,1 | Merge branch 'master' into 2.0.2 | 2.0.2 | ns_server | Status: MERGED | +2 | +1 |