777408 - pulling from http://hg.mozilla.org produces sporadic errors ("transaction abort! abort: 00changelog.i@887b70a4b6c2: unknown parent!")

Status: VERIFIED FIXED

(Developer Services :: General, task)

Description

[Jason Orendorff [:jorendorff]]

khuey observed it:

Kyle Huey@KYLEHUEY-PC /c/dev/mozilla-central
$ hg pull -u
pulling from http://hg.mozilla.org/mozilla-central
searching for changes
adding changesets
transaction abort!
rollback completed
abort: 00changelog.i@887b70a4b6c2: unknown parent!

Kyle Huey@KYLEHUEY-PC /c/dev/mozilla-central
$ hg pull -u
pulling from http://hg.mozilla.org/mozilla-central
searching for changes
adding changesets
transaction abort!
rollback completed
abort: 00changelog.i@887b70a4b6c2: unknown parent!

Kyle Huey@KYLEHUEY-PC /c/dev/mozilla-central
$ hg in
comparing with http://hg.mozilla.org/mozilla-central
searching for changes
abort: 00changelog.i@887b70a4b6c2: unknown parent!

Kyle Huey@KYLEHUEY-PC /c/dev/mozilla-central
$ hg out
comparing with http://hg.mozilla.org/mozilla-central
searching for changes
no changes found

It's also been observed by benjamin, tedshroyer, and jduell, with similar error messages (different revision numbers).

Comment 1

[Daniel Holbert [:dholbert]]

For anyone hitting this, stackoverflow suggests that "hg verify" might give more information:
 http://stackoverflow.com/questions/2733248/unknown-parent-error-in-mercurial

Comment 2

[ted shroyer]

I tried "hg verify" but got the same error after it ran.
(In reply to Daniel Holbert [:dholbert] from comment #1)
> For anyone hitting this, stackoverflow suggests that "hg verify" might give
> more information:
>  http://stackoverflow.com/questions/2733248/unknown-parent-error-in-mercurial

Comment 3

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

I saw the same error message on three different repos.  I think this is a server-side issue, not a client-side issue.

Comment 4

[Daniel Holbert [:dholbert]]

I believe (but I'm not 100% sure) that "hg verify" is entirely client-side -- so given that it produces the same abort (per comment 2), I think there's client-side bustage involved, too. 

(though the client-side bustage may have been directly caused by server-side bustage)

Comment 5

[Shyam Mani [:fox2mike]]

Corey, Ben - Did we push 2.2.3 anywhere?

Comment 6

[Ben Kero [:bkero]]

Yes, i pushed 2.2.3 to hgweb1 last night.  That node was also disabled in the load balancer.

Comment 7

[Shyam Mani [:fox2mike]]

Apparently not. So guys, we're looking for a timeline of when this started happening and is this related to http://bz.selenic.com/show_bug.cgi?id=3553 or is that a completely different issue?

Comment 8

[(no longer active)]

This is one of the symbols that I have been seeing in this bug: http://bz.selenic.com/show_bug.cgi?id=3553

Specifically see my last comment on that bug.

Comment 9

[(no longer active)]

And yes, this is a server side issue.  The error message is actually coming from the remote.

Comment 10

[(no longer active)]

Andreas was talking to me about how important my git mirror is yesterday, and my git mirror has been sort of broken since yesterday because of this bug...

Comment 11

[Shyam Mani [:fox2mike]]

Starting now, let's see when the next failure is? The only real tweak that was made to production yesterday was we enabled a cache with a 60 second TTL. That's now been turned off. I'd like to see if that was the cause for this.

Comment 12

[Corey Shields [:cshields]]

(In reply to Ehsan Akhgari [:ehsan] from comment #10)
> Andreas was talking to me about how important my git mirror is yesterday,
> and my git mirror has been sort of broken since yesterday because of this
> bug...

Can you explain this a bit?  Are you running your own git mirror that is being used for b2g?

Comment 13

[Ben Kero [:bkero]]

Shyam,

Zeus is the wrong place to be caching.  Mercurial supports its own caching, which I am speaking with the developers about using.

Comment 14

[(no longer active)]

(In reply to Corey Shields [:cshields] from comment #12)
> (In reply to Ehsan Akhgari [:ehsan] from comment #10)
> > Andreas was talking to me about how important my git mirror is yesterday,
> > and my git mirror has been sort of broken since yesterday because of this
> > bug...
> 
> Can you explain this a bit?  Are you running your own git mirror that is
> being used for b2g?

Yes.  I pull from m-c and inbound, convert stuff to git, and then push them to https://github.com/mozilla/mozilla-central and b2g uses that repository.

Comment 15

[(no longer active)]

For the benefit of people not reading our IRC conversation, now that IT has disabled the caching that was put in place yesterday, I'm testing this very aggressively to see if it happens again.

Comment 16

[Corey Shields [:cshields]]

This should be fixed, please reopen if that is not the case.

Comment 17

[(no longer active)]

It is definitely fixed.