[MDEV-24384] MariaDB crashing with long semaphore waits after upgrade from v10.2 to v10.5.8 Created: 2020-12-10  Updated: 2021-01-07  Resolved: 2021-01-07

Status: Closed
Project: MariaDB Server
Component/s: Storage Engine - InnoDB
Affects Version/s: 10.5.8
Fix Version/s: N/A

Type: Bug Priority: Critical
Reporter: Prabha Assignee: Unassigned
Resolution: Incomplete Votes: 0
Labels: need_feedback
Environment:

Linux server provisioned on Oracle cloud

Attachments: Zip Archive Mariadb.zip    
Issue Links:
Relates
relates to MDEV-24378 Crashes on Semaphore wait > 600 seconds Closed

 Description   

Hi,
Our prod database was crashing daily at least two times after we have upgraded to V10.5.8 from v10.2.

We disabled the adaptive hash index but it didn't prevent the crashes. Please do the needful.

 Comments   
Comment by Marko Mäkelä [ 2020-12-10 ]

In Mariadb.zip I see the following:

mariadb-10.5.8

 
OS WAIT ARRAY INFO: reservation count 22537437
 
--Thread 140242946348800 has waited at mtr0mtr.ic line 145 for 0.00 seconds the 
semaphore:
 
X-lock (wait_ex) on RW-latch at 0x7f90e41b1f78 created in file buf0buf.cc line 1299
 
a writer (thread id 140242946348800) has reserved it in mode  wait exclusive

There are no further references to that thread in the log output.

This looks similar to MDEV-24378 and could be duplicating MDEV-24188. Without having the stack traces of all threads, I am unable to diagnose either that or this report.

Generated at Thu Feb 08 09:29:35 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.