FYI, as I looked at the code that led up to this, I have realized that MySQL was doing a huge rollback, which ended up taking about five hours. It was rolling back about 2 million INSERTs, I think. The rollback really was not necessary, so I've changed the appropriate code so that it's no longer a transaction.
The culprit was some table locking that improved performance quite a bit when the tables were MyISAM. Gotta go look for more of those lurking in the corners, I guess. -- Nick Arnett Phone/fax: (408) 904-7198 [EMAIL PROTECTED] > -----Original Message----- > From: Heikki Tuuri [mailto:[EMAIL PROTECTED] > Sent: Tuesday, July 29, 2003 12:58 PM > To: [EMAIL PROTECTED] > Subject: Re: Really slow shutdown with Innodb, db not accessible? > Importance: High > > > Nick, > > ----- Original Message ----- > From: ""Nick Arnett"" <[EMAIL PROTECTED]> > Newsgroups: mailing.database.mysql > Sent: Tuesday, July 29, 2003 8:54 PM > Subject: Really slow shutdown with Innodb, db not accessible? > > > > For the last four hours or so, I've been waiting for MySQL > (4.0.12 on W2K) > > to complete a shutdown. The fast shutdown flag is not set > > (innodb_fast_shutdown=0), so I assume it is doing a purge and > merge... but > > in the meantime, I don't have any access to the server -- clients simply > > can't connect. This is a real problem, since it renders the database > > useless for a long period of time. My Innodb table is about 15 GB and > > probably has about 10 million records in various tables. > > > > When the darn thing finally shuts down, I'll restart with fast shutdown > on, > > but I'm wondering how foolish it would be to kill the process, > given that > > Innodb should then do a crash repair. Would the crash repair > take longer > > than what it's doing now? Would the server be inaccessible as > it is now? > > crash recovery is usually much faster than purge and merge. > > Killing the mysqld process is a legal (and the fastest :)) way of shutting > down InnoDB. > > Why did you set fast_shutdown=0? > > By the way, I am not sure the setting really affects the variable value at > all, since in versions < 4.0.15 there was a bug that it was specified as a > NO_ARG parameter. > > > Besides enabling fast shutdown, what else will help avoid this kind of > thing > > in the future? > > > > Thanks for any info... > > > > -- > > Nick Arnett > > Regards, > > Heikki > > > Phone/fax: (408) 904-7198 > > [EMAIL PROTECTED] > > > > -- > MySQL General Mailing List > For list archives: http://lists.mysql.com/mysql > To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED] -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
