[MXS-1961] Standalone master loses master status Created: 2018-07-06 Updated: 2020-08-25 Resolved: 2018-07-30 |
|
| Status: | Closed |
| Project: | MariaDB MaxScale |
| Component/s: | mariadbmon |
| Affects Version/s: | 2.2.11 |
| Fix Version/s: | 2.2.13 |
| Type: | Bug | Priority: | Major |
| Reporter: | markus makela | Assignee: | Esa Korhonen |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Description |
|
With a three-node setup of servers A, B and C with A being the initial master, a sequence of failovers from A → B → C will result in C being assigned as a standalone master. When B is restarted, C will lose the Master status and both B and C will end up in Running state. The expected behavior would be that C remains as the master and B rejoins as a slave. |
| Comments |
| Comment by Esa Korhonen [ 2018-07-24 ] |
|
Diagnosed and testing a fix. This bug happens in the case that there is a [Down] slave server (server3 in the test) which is replicating from the recently unblocked/returned alternative master (server2 in the test). The monitor then prefers server2. This can likely be fixed by disregarding downed slaves, although some tests needs to be ran to make sure this doesn't break other things. |