Details
-
Bug
-
Status: Closed (View Workflow)
-
Critical
-
Resolution: Fixed
-
10.3.13, 10.3.20
-
None
-
Centos 7
Description
We're experiencing following behavior seemingly randomly:
Mariadb 10.3 galera cluster with 3 nodes, restart one, and the node joins back to cluster normally (in less than a minute). But it immediately begins creating gcache.page.XXXXX file(s) after it's rejoined the cluster.
gcache size is 2 gigabytes and the db usage during the restart was quite small.
Afterwards, the same instance keeps creating new gcache.page.xxx files, while the other nodes do not.
And same thing when restarting other nodes, they rejoin normally and may or may not start to create gcache.page files right away.
What could be causing this kind of behaviour?
I believe the biggest "problem" this causes for us is that when a server node is using gcache.page files, it cannot serve IST to a joiner, but always does full SST, as I've understood.
Attachments
Issue Links
- relates to
-
MDEV-26968 Galera cluster gcache.page files creation at startup/restart (regression of MDEV-19084)
- Closed
-
MDEV-23060 Mariadb 10.4.12 gcache is created after every restart
- Open