These are chat archives for ManageIQ/manageiq/performance

17th
Feb 2018
Ron Valente
@rvalente
Feb 17 2018 00:16
Running into some swapping issues on my DB appliance (this is CloudForms 5.8.2.3) - roughly 8k managed objects, 1 provider. The replication process is the culprit as far as swap is concerned. All signs point to the max_wal_size being too small. Am I on the right track? - This manifested in the vmdb_accord failing to load in the UI when I was trying to see if vacuum was finishing. DB is 330GB in < 3 months.
Keenan Brock
@kbrock
Feb 17 2018 00:36
@dmetzger57 purging PR ManageIQ/manageiq#17017 /cc @Fryguy
@rvalente looks like there are some tables that are not being cleaned up
@rvalente replication questions better asked in core or main room. Also cc @carbonin
Ron Valente
@rvalente
Feb 17 2018 00:39
@kbrock I'll take a look. Tuning the max_wal_size to 4GB and wal_buffers to 32MB has dropped my swap to 2GB, and silenced the replication issues with checkpoints updating too quickly
Keenan Brock
@kbrock
Feb 17 2018 00:39
nice
may want to run https://github.com/kbrock/bin/blob/master/pdu -- pdu -l 40 -i vmdb_production
to see where the extra space is bloating
(-i option may be too much info - it shows indexes)
book reading time for the kids - will try and circle back in a bit
Ron Valente
@rvalente
Feb 17 2018 00:42
Ciao, Thanks for the help!
Keenan Brock
@kbrock
Feb 17 2018 01:25
@rvalente did you figure out the largest of your objects?
Ron Valente
@rvalente
Feb 17 2018 10:22
@kbrock metrics_17 -> metrics_23 are all 16GB - 35GB.