Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Incomplete
-
10.6.7, 10.6.9
-
#server1 - CentOS 7.9, DirectAdmin 1.642, MariaDB 10.6.9
#server2 - CentOS 8.5 (AlmaLinux 8.5), DirectAdmin 1.642, MariaDB 10.6.7
Description
Hi,
We have over 2000 databases (innodb)
In April 2022 we moved from #server 1 to #server 2 (which is few times faster than #server1) and we have some troubles:
1. On #server1 there was mariadb 10.4.x and any backups/upgrades were creating really fast (30-60 seconds)
2. We uploaded backups to #server2 but installed there MariaDB 10.6.7 and.... dumping databases is taking ~60-70 minutes + after for example do command on SSH like "service mariadb restart" is making some crash recovery etc.... which is taking 1-2 hours also.... on #server1 with mariadb 10.4.x there wasn't any problems like that, I could restart mariadb anytime and there wasn't any problems...
3. Today (20.08.2022) I logged to #server1 and thinking "hmmm maybe i'll upgrade mariadb 10.4.x to 10.6.x to check if anything will happened like in #server2" - and what? If there were mariadb 10.4.x mysqldump make dumps in 30-60seconds (2100 databases).... after successfully upgrade to mariadb 10.6.x, I tried to reinstall mariadb from 10.6.9 to 10.6.9 (to check how fast is mysqldump) and that stupid mariadb 10.6.9 is doing mysqldump like 1 database every 2 seconds...so 2100 databases give us 60-70 minutes - why it's so much slower than 10.4.x? What should I do? It's terrible problem for us and I don't know how to fix that.... I'm also scared to upgrade mariadb 10.6.7 to 10.6.9 on #server2 (production server) because of recovery crashes after restart etc.
Any ideas??
Attachments
Issue Links
- relates to
-
MDEV-29967 innodb_read_ahead_threshold (linear read-ahead) does not work
- Closed
-
MDEV-25417 reduce InnoDB buffer pool load time
- Closed
-
MDEV-28909 Write performance not scale to NVMe SSD
- Stalled