Odd RT lockup

This is more for sites with a large number of users directly using RT,
which the RIPE NCC does not have. After putting another section onto the
wonder that is RT, I received a complaint that the system ‘wasn’t
working’. After trying to log in or display the queue from the command
line and finding no response, I looked on the SQL server to find a silly
number of mysql processes around, and the number of threads in slow-query
mode increasing.

As we don’t log individual sql queries on that box, I’m unable to say
which query blocked originally. It differs from previous hiccups in that
it had been locked solid for ‘some time’. I presume that it is the old
table locking problem, which will be fixed when we upgrade that SQL
cluster to the latest and greatest.

Do any larger (in number of users) RT sites want to comment on their
solutions to the lock problem?

Jesse, is there magic in DBIx::SearchBuilder that I’m missing in order to
set timeouts ?

One that I should follow up on is splitting the reads from the writes as
per previous indications.

                         Bruce Campbell                            RIPE
               Systems/Network Engineer                             NCC
             www.ripe.net - PGP562C8B1B                      Operations