Re: Sporadic Query "Freeze" with InnoDB, highly concurrent workload

[Steve Ramage]

Hmmmmm so it has happened again, and what is odd is that I have killed all the requesting tasks but not the queries in MySQL and it is still going. They were killed an hour before I took the snapshots. I actually think it must be some kind of live lock. The only things that pop out at me are that I wonder if setting the isolation level to READ_COMMITTED would help, instead of REPEATABLE_READ. Any ideas as to why no queries seem to be making forward progress. Also CPU usage seems to be very high for the MySQL process (~600%).

Eventually I went to bed and got up again, and everything had cleared out. When it gets back to a steady state the CPU usage is like 2% normally.

INNODB Mutex Status: http://www.cs.ubc.ca/~seramage/innodbmutex.txt
Global Status: http://www.cs.ubc.ca/~seramage/globalstatus.txt
INNODB Engine Innodb Status: http://www.cs.ubc.ca/~seramage/innodbstatus.txt
InnoDbTrx: http://www.cs.ubc.ca/~seramage/informationSchemaInnodbTrx.txt
Process List: http://www.cs.ubc.ca/~seramage/processlist.txt