"User timeout caused connection failure." viewing branches

[William Grant]

One of my Launchpad-hosted branches says "User timeout caused connection failure." This doesn't make sense; I didn't time out, nor should it have to connect to anything.

[Jonathan Lange (jml)]

This is the error displayed when the branch mirroring process (the puller worker) doesn't make any progress for a long period of time. The error could be avoided with a higher timeout or better progress reporting from Bazaar. The error text itself comes straight from Twisted and isn't relevant here. Improving the text would be a trivial fix.

[Michael Hudson-Doyle (mwhudson)]

A link to the branch would help :)

[William Grant (wgrant)]

Sorry, I'd suspected that it would be an intentional and obvious lp-bzr error message, so you would have recognised what it meant immediately (or been able to grep for it). I see now that it was not from your code. As discussed with jml, the branch in question is https://code.launchpad.net/~wgrant/ubuntu-cve-tracker/mainold.

[Jonathan Lange (jml)]

That error message is Twisted's twisted.internet.error.TimeoutError. Here, it probably means that the process that was actually mirroring the branch stopped sending any information, was treated as non-responsive and then killed.

We could probably change the visible message quite easily, although it will be nigh impossible to verify the fix on a production environment.

Note that now this will never happen for hosted branches, since they are no longer mirrored internally.

[Ursula Junque (ursinha)]

I found some oopses related to this: OOPS-1861CIW47, OOPS-1861CIW77. I couldn't reproduce the error with the branches in the oopses, but could consistently with wgrant's branch. Is this a real failure or not really an OOPS? If not an OOPS, can we log them as informational or don't log them at all?

[Robert Collins (lifeless)]

It is a real failure; we should see about gathering more diagnostics.