Lance Albertson wrote:
kashani wrote:
Chris Bainbridge wrote:
I'm curious what the problem is with bugzilla and it's db
interactions? You're suggesting a specific issue rather than general
db performance issues like fs, io scheduling, raid1, hyperthreads,
etc.?
It's most likely related to Bugzilla using MyISAM tables by default
which is fine in a small environment. However as concurrency grows along
with table size those full table locks begin to cause issues.
Additionally most www-apps are not known for their well thought out db
schema. Performance of the underlying hardware plays a part, but can be
overshadowed pretty quickly by query and table inefficiencies.
The usual fix without touching the internals of the software is
doing master/slave replication and allowing the selects to happen on the
slave and writes on the master. However you would need to change most of
the queries to point to the slave server which is usually not trivial.
It sounds like this is the path the Infra team is pursuing.
Thanks a lot for this explanation.
1++
You got it right :-)
I'm not out to blame anybody, but if infra had communicated what the problem
exactly is once they found it out, you wouldn't have ended up with all those
"I'm sick and tired of your "we're working on it"". Asking people for patience
is easy, but it's hard to swallow when you don't understand what the problem is
at all.
--
Kind Regards,
Simon Stelling
Gentoo/AMD64 Developer
--
gentoo-dev@gentoo.org mailing list
