Recently we are suffering table locks while backup is running.
Before the backup is sent we change wsrep_desync=TRUE, but we think that is not working fine.
Other way when backup node is desynced is receiving data from the cluster, and if we change somethin on it we can see the change in the cluster. I thought that souldn’t be this way.
We use percona-xtradb-cluster-server-5.7 and percona-xtrabackup-24
As explained on the official doc MySQL wsrep Options — Galera Cluster Documentation
wsrep_desync option will prevent the node from triggering flow control but replicated events will still be applied on the node
But that I don’t understand is if desynced node sends data to the other nodes.
As I’m seeing there are locks in te desynced node that are arriving to the other nodes of the cluster.
We fear that backup locks could be propagating to the synced nodes