Hello,
starting with an explanation first... i have some sort of 'cold
backup' system: it rsyncs a postgres dump (size: ~1.5TB) from the main
system (created with 'pg_dump $dbname > db.sql) to the backup system
(10TB hdds in a raid10), and then automatically imports that dump
(psql $dbname < $db.sql) into postgres. So the backup system
stores/provides the database of each day (cleanup of older stuff done
manually). copy+import normally finished within 16-18 hours.
About 4 weeks ago both systems were running with postgres12 on debian
buster; then the switch to postgres14 was done, first on main system,
later on the backup system (both with pg_upgrade, no problems here).
But after that switch the import time on the backup system increased,
now taking more than 24 hours. A dist-upgrade to debian bullseye was
done, but no improvement.
I was checking release notes (releases 13+14), but, somehow, nothing
catched my eye.
hardware of backup system:
128G ram, raid10 (ext4, 4 disks) for postgresql, raid10 (ext4, 6
disks) for sql backup files
shared_buffers = 32768MB
work_mem = 16384MB
maintenance_work_mem = 16384MB
effective_cache_size = 16GB
no read/write operations on the backup system. any idea what could've
changed? more information needed?
thx,
Hubert
