Page 1 of 1
another distraction
Posted:
Sat Nov 01, 2003 8:41 pmby josh
Do any of you have much experience with mysql? We're running v 3.23 -- this was part of the standard buildout of the ISP who set up the server, and I'm thinking they believe it to be stable.
I'm seeing the big tables get corrupted more and more -- I have to run the myisamchk repair tool to fix them. So far they've been recoverable, but it is worrisome, and occasionally causes a crash (btw, spameater now dies with a TEMPFAIL error code for sendmail, so it's treated as a temporary failure).
The "big" tables aren't that big -- Emails has about 390k records, and Users has about 43k records. They do get positively slammed with updates though.
Any tips on stabilizing this? Upgrading? Switching (PostGres?)?
I am researching that
Posted:
Sun Nov 02, 2003 3:59 amby SysKoll
I am researching that. I wasn't aware of a MySQL table corruption issue. That's bad. Short of switching to DB2, there must be a way to get a stable DB that can withstand massive updates. PostGreSQL looks like a possibility.
Test case procedure for MySQL
Posted:
Sun Nov 02, 2003 9:02 pmby SysKoll
Josh,
The MySQL developers have a nice little doc outlining how to create a test case and report a bug. It requires making a copy of the tables. Since you need to stop mysqld to run the myisamchk program, maybe you can use that opportunity to make a copy of the corrupt table and run that problem isolation procedure? (You
do stop the mysqld daemon when you run the repair utility, right?)
http://www.mysql.com/doc/en/Reproduceable_test_case.html
What error does myisamchk give you?
Posted:
Sun Nov 02, 2003 9:06 pmby SysKoll
Josh,
I forgot to ask the most obvious question: When you run myisamchk, what messages do you see? What kind of error do you get?
Please let us know where the running code is on the server so we can take a look at it.
Posted:
Wed Nov 05, 2003 2:12 amby josh
Sorry about the dead air recently -- slammed at the day job again.
I didn't record the myisamchk output last time, but I do have this from the log:
read_const: Got error 127 when reading table ./XXXXX/Users
(where XXXXX is the name of the database). Not real specific, I know.
Posted:
Fri Feb 20, 2004 4:15 pmby maratheamit
Josh, how is mysql holding up these days? Do you still see table corruption on a regular basis?
Posted:
Fri Feb 20, 2004 5:28 pmby josh
no - I haven't seen the problem for awhile...
I'm wondering: awhile back I had tried to set up a test version of the website using Apache::Registry with persistent database connections. I *think* I probably compiled the apached mod against the wrong set of mysql libs. I know that whenever I accessed the site, the test database went to hell. Maybe that was what behind the problems we saw earlier? Don't know.