[MDEV-24636] InnoDB: Semaphore wait has lasted > 600 seconds. We intentionally crash the server because it appears to be hung. Created: 2021-01-21  Updated: 2023-05-12  Resolved: 2023-05-12

Status: Closed
Project: MariaDB Server
Component/s: Storage Engine - InnoDB
Affects Version/s: 10.3.16
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: chencheng Assignee: Marko Mäkelä
Resolution: Incomplete Votes: 0
Labels: crash, semaphore
Environment:

MariaDB Version: 10.3.16
system: centos7 (3.10.0-123.20.1.el7.x86_64)
cpu: 4 cores or 8 cores
hardware: Alibaba Clould ECS

Attachments: Text File 4C.txt     Text File 8C.txt     File server.cnf    

 Description   

“--Thread 140014490990336 has waited at srv0srv.cc line 915 for 921.00 seconds the semaphore:
Mutex at 0x7f5852701310, Mutex SRV_SYS created srv0srv.cc:1020, lock var 2”

Our MariaDB server often restart due to the lock. And it happened even every hour after upgrading the CPU from 4 cores to 8 cores.

We guess that has something to do with adaptive hash index. Bug it didn't work after set innodb_adaptive_hash_index=0.

We have no idea. Can we solve this problem by upgrade to a newer version?

Environment->
MariaDB Version: 10.3.16
system: centos7 (3.10.0-123.20.1.el7.x86_64)
cpu: 4 cores or 8 cores
hardware: Alibaba Clould ECS

 Comments   
Comment by Marko Mäkelä [ 2023-04-14 ]

For diagnosing any hangs, we’d need full stack traces of all threads during the hang. Can you provide that?

There have been several fixes to the adaptive hash index since this release. One prominent change was MDEV-22456.

Generated at Thu Feb 08 09:31:30 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.