I recently upgraded a customers server from a P3/XEON box to a 64-bit Opteron with 16GB RAM. They have more data than can easily be handled by a 32-bit machine with the inherent limit of ~2GB on innodb_buffer_pool and is now much faster. The old server ran 4.0.18 and I upgraded to 4.0.20 on the new one. No problems on the old server other than it being slower. Database "freezes" on the new server though.
RedHat Enterprise 3WS, fully up2date. MySQL binary RPMs for AMD64 won't even start for me, what were they built on?. I installed the .src.rpm and built one myself (-bb --target amd64). Installed that and all seems well. Our application occasionally will lock a table for writing, do an insert or update, then unlock the table. The customer calls me with all clients "froze". A 'show processlist' shows a table in a state of locked, I see a process with the lock command still there attempting to lock it, and other accesses to it all stalled/frozen as well. I issue an 'unlock tables' command and sometimes it helps, usually not. I can sometimes kill the id of the lock command itself (which should have ran and _not_ been there anymore but it's _hung_ as well) and that will help. Sometimes 'unlock tables' a number of times will help. About 40% of the time so far I've had to restart MySQL entirely to free it up for them though. Once or twice this has only lasted 30 seconds before it's right back in the same boat and I restart it again. I've made my.cnf changes 3 or 4 times and each time thought I had resolved things since most of a day would pass with no problems. Then all of a sudden, 3 freezes in an hour. So, right now I'm compiling 4.0.18 and will downgrade to that and see what happens. Here's an error I'm getting in the log quite a lot and never saw with 18: 040610 18:17:17 Aborted connection 22 to db: 'dbname' user: 'username' host: `win_pc_oddb_connected' (Got an error reading communication packets) At this point I think this is a bug with 4.0.20 and and will reply with my results with .18 after a few days if there's an interest. As an aside, I have 3 other Opteron servers with similar specs, one at this site and no problems on them. The 2nd box at this site is a replication slave to the first one though and does nothing but that, sitting idle as a spare in case the first were to break (I'll tinker with the cluster product one of these days). It binary logs updates and sends them up to a web server with similar specs. The server with lots of Windows/ODBC clients and most of the activity is the one that freezes, no such problem on any of the others. Their use is very much different though and I suspect that's a key to the problem. -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
