patchwork.ozlabs.org down, and e-mails not recorded
dja at axtens.net
Sat Feb 17 01:24:46 AEDT 2018
Andrew mentioned to me that we might have a race condition, and so I had
a quick poke at loading patches in parallel. Sure enough I was able to
easily hit an issue at least on the MySQL backend. (OzLabs uses postgres
but it's still a good data-point as it shows we're not handling this
sort of thing in code.)
This is quite possibly related to your issue - I will do some more
investigation and let you know how we go.
> On Wed, 14 Feb 2018 14:56:56 +1100, Andrew Donnellan wrote:
>> > Can we do something about this ? We're really happy otherwise by the
>> > patchwork instance at ozlabs.org, and we would hate having to run our
>> > own instance :-/
>> We've done some digging...
>> These are the other two patches in the series, and yet patchwork has
>> assigned them two different series, so evidently something screwy is
>> happening there.
>> sfr took a look at the mail logs for us, it looks like:
>> - all 3 emails were received and processed by our SMTP system
>> - patch 2 was received first, and patches 1 and 3 were received during
>> the same second a couple of seconds later
>> Patch 3 has not hit the database at all, it's possible we've hit some
>> race condition somewhere that prevented it from being saved at the same
>> time as patch 1, or perhaps we ran into something completely
>> different... we're going to add some extra logging to see if we can
>> capture more info next time this happens.
>> Many apologies for the inconvenience!
> As asked by Jeremy, I've changed the patchwork e-mail address to which
> the Buildroot patches are being sent, so that you can add more logging,
> but only for Buildroot stuff.
> Regarding the patches not properly assigned to series, we have already
> seen that many times, even in situations were no patches are lost. Some
> patches in the series are properly marked as being part of the series,
> and some are not. It feels like if some patches are received before the
> cover letter, patchwork gets confused, but that's just a guess. It
> doesn't happen very often, but it does happen. However, this is a lot
> less annoying that "losing" patches.
> Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons)
> Embedded Linux and Kernel engineering
> Patchwork mailing list
> Patchwork at lists.ozlabs.org
More information about the Patchwork