MariaDB deployment does not respect quorum and may break cluster
Bug #1859145 reported by
Radosław Piliszek
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
kolla-ansible |
Fix Released
|
High
|
Radosław Piliszek | ||
Rocky |
Won't Fix
|
High
|
Unassigned | ||
Stein |
Won't Fix
|
High
|
Unassigned | ||
Train |
Fix Released
|
High
|
Radosław Piliszek | ||
Ussuri |
Fix Released
|
High
|
Radosław Piliszek |
Bug Description
Current code does not wait for MariaDB to recalculate quorum when starting and stopping MariaDB containers. This may lead to WSREP issues (and failures) and in the worst case require a recovery.
This most notably affects upgrades of MariaDB clusters but also its reconfigurations (due to user action or our change). Hence high importance.
Changed in kolla-ansible: | |
status: | Triaged → In Progress |
To post a comment you must log in.
The fix is to separate bootstrap, deployment of new members and restart of old members, ensuring restart uses 3 phases (aka batches) not to break quorum.