[rt-devel] Re: RT SQL 'lockup' (mysql->99% cpu)

Jesse Vincent jesse at bestpractical.com
Fri Jul 25 12:11:14 EDT 2003



On Fri, Jul 25, 2003 at 01:10:24PM +0100, Iain Price wrote:
> > > This is with RT 3.0.2, MySQL 3.23.55 Apache 1.3.28 and mod_perl 1.28
> >
> > First up, move up to mysql 4.0.13 or newer. really. It's internal query
> > processor is WORLDS better. It won't cure cancer, but it should
> > definitely help performance. Especially if you turn on query caching.

> I dont mean to doubt your views or knowledge on mysql etc but given how
> little use mysql4 over mysql3 is to me (an uncompletable query still is...)
> should i spend a day looking at query caching and performance (remember this
> is works time not mine ;) or are you just suggesting any ideas you can think
> of rather than knowing that this will solve my issue.

It's not a magic bullet that will magically make all bogus queries
perform in negligable time, but its query processor has proven to be
much more advanced and to NOT DIE in situations where Mysql 3.23 curls
up into a little ball and refuses to deal with the world. I can't say
whether you should spend a day of your work time on it. That would
depend on how much of a problem this issue is for your organization.


> I've become a firm believer the only solution is to better the generated
> query or prevent users generating this particular search, since optimising
> mysql seems to lead nowhere (except to even more impossible queries) and
> having a user base that is 100% competent all the time with every entry
> seems a pipe dream... You should see some of the e-mail addresses/typos they
> type into RT (omg!), and i get to sort the bounces :S
> 

Great. I don't know when we're going to get to this query builderchange internally.
If it's an issue that's important to you, as it appears to be, I'd love to
see a patch.

	Jesse

> Iain
> 

-- 
http://www.bestpractical.com/rt  -- Trouble Ticketing. Free.



More information about the Rt-devel mailing list