Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Not a Bug
-
None
-
None
-
single node Centos 7.8
Description
Reported by customer:
this morning we found a lot of queries that normally take aprox 1 second, waiting for more than 6000 seconds.
We tried to kill the queries from mariadb shell, but after 5 minutes they all where yet there, despite the status was "killed".
Also tried to cleanly shutdown mariadb process (systemctl stop mariadb) but after several minutes the process was yet alive.
We had to kill mariadbd with kill -9
The queries running where all involving Columnstore tables.
We are using columnstore_select_handler=0 as GLOBAL because we need to do INSERT....SELECT and several kind of
joins that select_handler still not support (after years, that's incredibile), but we don't know if this setting is causing this issue or any other in terms of performances.
The server is a single node with both InnoDB and Columnstore.
In case report:
- snapshot of sqlyog showing query stalled
- snapshot of sqlyog showing query killed but still there
- output of "/usr/bin/columnstoreSupport -a";
- mariadb error.log (sw-mcs.err)
- output of "SHOW STATUS"