[Patchew-devel] [PATCH 1/1] debug: Enable debug toolbar for superuser in production deploy

Fam Zheng famz at redhat.com
Wed Oct 31 09:42:28 UTC 2018


On Wed, Oct 31, 2018 at 5:24 PM Paolo Bonzini <pbonzini at redhat.com> wrote:
>
> On 31/10/2018 10:19, Fam Zheng wrote:
> > On Wed, Oct 31, 2018 at 4:58 PM Paolo Bonzini <pbonzini at redhat.com> wrote:
> >>
> >> On 31/10/2018 04:09, Fam Zheng wrote:
> >>>>> Sounds good, I'll tweak it and repost.
> >>>> Go ahead and deploy it. :)
> >>> The problem is that adding the middleware slows down tests. I'll see if this can
> >>> be improved.
> >>
> >> How did you measure this (and what was the result)?
> >
> > Just 'time ./manage.py test tests'. I think it is slowed by 2x or
> > something, which I don't like given that it's already a couple
> > minutes.
>
> Oh. :(
>
> >> Alternatively, perhaps you could copy the database over to
> >> next.patchew.org and only turn it on for next.patchew.org.  This at
> >> least lets us look at "explain" results with realistic data.
> >
> > I'd like to avoid differentiating next.patchew.org and patchew.org in
> > terms of alive code. Maybe analyzing the live DB locally is okay too.
> > I'll try that first and share the data or result with you.
>
> Yes, please share the up-to-date database with me.  My hope is that it's
> just a couple indices that are missing.  In particular, I suspect that
> we need indices that cover both foreign keys and date, otherwise SQLite
> won't optimize for pagination and will first scan the entire database
> *then* sort.

Sure, I'll share with you the DB file offline.

Fam

>
> Paolo
>
> > I'm also counting on switching to postgres to improve performance, too.
> >
> > What do you think?
> >
> > Fam
> >
> >>
> >> Paolo
>




More information about the Patchew-devel mailing list