launchpad-dev team mailing list archive

[Robert Collins <robert.collins@xxxxxxxxxxxxx>]

I just had a catchup with Stuart about this specific aspect of the
change to pg 8.4.

The current status is:
 - many queries are faster
 - some queries have regressed massively on staging and on prod
 - some queries have regressed only on prod

As a result, Stuarts recommending that we get a fresh OOPS for any
timeout bugs filed before we were completely on pg 8.4 - some older
bugs may be fixed now by ppostgresql improvements.

He is currently trying to identify why some things are fast on staging
and slow on production - I'm going to file a bug in a minute for this,
because if staging is faster than prod at (well anything) it stops
being a good predictor of production performance, and we need that
predictive ability to avoid panics.

I'm going to add a tag 'pg83' to all the existing timeout bugs; if
you're looking at fixing a timeout bug with that tag, please find a
fresh OOPS for it and remove the tag. If you can't find an OOPS, and
the urls that were slow are consistently fast, I recommend you
consider closing the bug (with a note that it may need to be
reopened).

That said, the OOPS system is currently a little sick - its not
finding OOPS from lpnet, and Ursula couldn't see why; hopefully Diogo
can look at it in 7-8 hours from now.

Cheers,
Rob