We have a development database server with Percona-server 5.7.15-9 on it. It replicates two schemas from two different production servers using multi-source GTID replication. Lets call these schemas
On the dev server we clone this replicas of production databases to get databases for development. They are called
2_bob, etc. All of them use the same instance of MySQL.
For fast cloning we use Percona XtraBackup as described here https://www.percona.com/doc/percona-xtrabackup/2.4/innobackupex/restoring_individual_tables_ibk.html
In the past there were only one replica (
alice) and we used binary log position for replication instead of GTID. These times everything worked fine and fast. One day (I'm not sure when exactly) it became broken.
Now when I perform
ALTER TABLE 2_alice.access_group IMPORT TABLESPACE query, it hangs in the 'System lock' state. And could be hanging in this state from 1 min 'till 1 hour and more (then it works). There are no more active connections instead of two replicas, but they doesn't use
IMPORT TABLESPACE query hanging and how could I debug this case?