launchpad-dev team mailing list archive

[Deryck Hodge <deryck.hodge@xxxxxxxxxxxxx>]

Hi, Robert.

On Tue, Aug 24, 2010 at 2:20 PM, Robert Collins
<robert.collins@xxxxxxxxxxxxx> wrote:
> Theres a 40 second gap between a fairly innocuous call:
> 486.    5408    4ms     launchpad-main-master   UPDATE Bug SET
> heat=calculate_bug_heat(622129), heat_last_updated=CURRENT_TIMESTAMP
> AT TIME ZONE 'UTC' WHERE Bug.id = %s
> and the end of request glue:
> 487.    45213   2ms     session UPDATE SessionData SET last_accessed =
> CURRENT_TIMESTAMP WHERE client_id = %s AND last_accessed <
> CURRENT_TIMESTAMP - '540 seconds'::interval
>
> This action sends email, so its a pretty good bet that that is it. But
> - how many emails is it sending? Aren't emails from bugs meant to be
> deferred *anyway* ?
>

Perhaps this was rhetorical to highlight the problem you see here (and
sorry I'm late in responding to this), but to answer the general
question about email:

Emails that are batched are sent out of app, which is most email
generated by a bug.  Any change to the bugtask, adding a comment,
changing an attribute of the bug itself (summary, description),
adding/removing tags, and so on.  Subscriptions are the main
difference where the email is sent from the web app notifying the user
that she has been subscribed (if subscribed by someone else).  This
will also re-calculate heat, since changes in subscription count
affect heat.  You could subscribe a team to a bug and sending the
mails could take some time.

I don't think we should do this.  In fact, I've always felt sending
email from a web app was one of those thou-shalt-not kind of things,
but I haven't been able to get this fix scheduled for malone.  To
reach our performance goals, we will most certainly have to change
this.

Cheers,
deryck


-- 
Deryck Hodge
https://launchpad.net/~deryck
http://www.devurandom.org/