maria-discuss team mailing list archive
-
maria-discuss team
-
Mailing list archive
-
Message #06501
Re: Mariadb 10.4 to 10.6 performance regression
-
To:
<maria-discuss@xxxxxxxxxxxxxxxxxxx>
-
From:
"Reinis Rozitis" <r@xxxxxxx>
-
Date:
Mon, 30 Jan 2023 20:36:32 +0200
-
In-reply-to:
<000301d934b9$1e4da760$5ae8f620$@roze.lv>
-
Thread-index:
AQHB6KAldDkAEEUNxrHL5ToNPXXDea7l3XmA
> upgrade from 10.4.26 to 10.6.11, but the performance difference is so
significant/abysmal
After some time to see if maybe the Innodb buffer pool warming up might help
(it didn't) downgraded to 10.5.18 and the performance seems to be
immediately on same level as with 10.4.
With 10.6 the slave produced/logged quite more slow queries of which
majority were just COMMITS (never seen on 10.4 or 10.5):
# User@Host: [SQL_SLAVE] @ []
# Thread_id: 2044 Schema: db QC_hit: No
# Query_time: 1.913520 Lock_time: 0.000006 Rows_sent: 0 Rows_examined: 0
# Rows_affected: 0 Bytes_sent: 0
SET timestamp=1674989432;
COMMIT;
Any ideas/comments?
Does it make sense to test with any short-term branches? 10.10+ ?
wbr
rr
Attachment:
downgrade_to_10.5.png
Description: PNG image
References