Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Incomplete
-
None
-
RHEL 7, kubernetes environment using Helm
Description
When deploying a galera cluster in a kubernetes container environment (via helm), the likelyhood of an entire cluster failure increases significantly. The ability to manually resolve issues where galera fails to come up due to possible loss of transactions is limited.
Need ability (via option?) for galera to automatically attempt to come up even in the case where data loss could occur. The conditions that I have seen so far are:
1. grastate.dat safe_to_bootstrap is 0 (which we have already automatically set to 10
2. Galera fails to come up and requires mysqld --user=mysql --tc-heuristic-recover rollback to be run to come up
In kubernetes container environment things need to be more automatic