Mailing List Archive

wiki.zope.org outage
Hi all. I apologise for today's outage of wiki.zope.org and zwiki.org .
Things were working ok until about lunchtime.. I and others even were doing
a bunch of maintenance. Around noon I packed the zodb (from 3 -> 2G). This
afternoon I notice zope down and I can't start the zeo client with either
the old or new Data.fs. After connecting to the server it says:

2007-02-14 17:51:50 BLATHER ZEO.zrpc (17346) CT: exiting thread:
Connect([(2, ('localhost', 8888))])
2007-02-14 17:51:51 ERROR ZODB.serialize Unpickling error:
'(cApp.ApplicationManager\nApplicationManager\nq\x$
Traceback (most recent call last):
File "/zope-2.9.4/lib/python/ZODB/serialize.py", line 597, in getState
return unpickler.load()
EOFError

I would appreciate any ideas at all. Best,

-Simon


_______________________________________________
Zope-web maillist - Zope-web@zope.org
http://mail.zope.org/mailman/listinfo/zope-web
Re: wiki.zope.org outage [ In reply to ]
On Feb 14, 2007, at 8:56 PM, Simon Michael wrote:

> Hi all. I apologise for today's outage of wiki.zope.org and
> zwiki.org . Things were working ok until about lunchtime.. I and
> others even were doing a bunch of maintenance. Around noon I packed
> the zodb (from 3 -> 2G). This afternoon I notice zope down and I
> can't start the zeo client with either the old or new Data.fs.
> After connecting to the server it says:
>
> 2007-02-14 17:51:50 BLATHER ZEO.zrpc (17346) CT: exiting thread:
> Connect([(2, ('localhost', 8888))])
> 2007-02-14 17:51:51 ERROR ZODB.serialize Unpickling error:
> '(cApp.ApplicationManager\nApplicationManager\nq\x$
> Traceback (most recent call last):
> File "/zope-2.9.4/lib/python/ZODB/serialize.py", line 597, in
> getState
> return unpickler.load()
> EOFError
>
> I would appreciate any ideas at all. Best,

Do you have ZEO set to use a persistent cache? If so, remove it and
try again.
If not:

It looks like your database is corrupted. Have you tried using
the .old file
left by the pack? If that doesn't work, you'll probably need to use
a backup.

Otherwise, it might be possible to replace this object. I don't
think ApplicationManagers have any precious state. They really
shouldn't be persistent at all IMO. You might be able to use zopectl
debug to replace this object with a new instance.

Jim

--
Jim Fulton mailto:jim@zope.com Python Powered!
CTO (540) 361-1714 http://www.python.org
Zope Corporation http://www.zope.com http://www.zope.org



_______________________________________________
Zope-web maillist - Zope-web@zope.org
http://mail.zope.org/mailman/listinfo/zope-web
Re: wiki.zope.org outage [ In reply to ]
Jim Fulton wrote:
>> '(cApp.ApplicationManager\nApplicationManager\nq\x$

Jim, thanks for the tips. I'm sorry, I see now the above was truncated. For
the record a more complete error is below.

fstest.py reported no problems and today I found ZODB/fsrecover.py (not in
SOFTWARE_HOME/bin), which processed the whole zodb with "0 bytes removed
during recovery".

I tried a backup today. I could swear it showed the same problem, and I got
it to go away removing var/{pts,LatexWiki}/* and Products/*, but I can't
reproduce either of these; anyway, the long and short of it is the server
is back up and running, and any changes since 2007/02/14 04:19 MST will
have to be redone.

Thanks all. My next two projects are to export to a fresh zodb to reduce
the size, and to upgrade to 2.10.x.

-Simon

2007-02-15 18:18:42 ERROR ZODB.serialize Unpickling error:
'(cApp.ApplicationManager\nApplicationManager\nq\x01)tq\x02.}q\x03(U\x12TranslationServiceq\x04U\x08\x00\x00\x00\x00\x00\x06D\x11q\x05QU\x08_objectsq\x06(}q\x07(U\tmeta_typeq\x08U\x13Database
Managementq\tU\x02idq\nU\x08Databaseq\x0bu}q\x0c(h\x08U\x12Version
Managementq\rh\nU\x08Versionsq\x0eu}q\x0f(h\x08U\x13WebDAV Lock
Managerq\x10h\nU\x08DavLocksq\x11u}q\x12(h\x08U\x12Product
Managementq\x13h\nU\x08Productsq\x14u}q\x15(h\x08U\x11Debug
Informationq\x16h\nU\tDebugInfoq\x17u}q\x18(U\tmeta_typeq\x19U\x12Wing
Debug
Serviceq\x1aU\x02idq\x1bU\x10WingDebugServiceq\x1cu}q\x1d(U\tmeta_typeq\x1eU\x1dPlaceless
Translation Serviceq\x1fU\x02idq
h\x04utq!h\x1cU\x08\x00\x00\x00\x00\x00\x05i\x15q"QU\x08Productsq#U\x08\x00\x00\x00\x00\x00\x00\x00\x04q$Qu.'
Traceback (most recent call last):
File "/zope-2.9.4/lib/python/ZODB/serialize.py", line 597, in getState
return unpickler.load()
EOFError

_______________________________________________
Zope-web maillist - Zope-web@zope.org
http://mail.zope.org/mailman/listinfo/zope-web
Re: Re: wiki.zope.org outage [ In reply to ]
On Feb 15, 2007, at 3:44 PM, Simon Michael wrote:

> Jim Fulton wrote:
>>> '(cApp.ApplicationManager\nApplicationManager\nq\x$
>
> Jim, thanks for the tips. I'm sorry, I see now the above was
> truncated. For the record a more complete error is below.
>
> fstest.py reported no problems and today I found ZODB/fsrecover.py
> (not in SOFTWARE_HOME/bin), which processed the whole zodb with "0
> bytes removed during recovery".

Note that both of these analyze the database structure but not the
pickle contents. It would be nice to have a tool that also analyzed
the pickles.


> I tried a backup today. I could swear it showed the same problem,
> and I got it to go away removing var/{pts,LatexWiki}/* and Products/
> *, but I can't reproduce either of these; anyway, the long and
> short of it is the server is back up and running, and any changes
> since 2007/02/14 04:19 MST will have to be redone.

Did you try the .old file left by the pack?

> Thanks all. My next two projects are to export to a fresh zodb to
> reduce the size, and to upgrade to 2.10.x.

The pack should work. I'd be very surprised if the problem was
caused by the pack.

Jim

--
Jim Fulton mailto:jim@zope.com Python Powered!
CTO (540) 361-1714 http://www.python.org
Zope Corporation http://www.zope.com http://www.zope.org



_______________________________________________
Zope-web maillist - Zope-web@zope.org
http://mail.zope.org/mailman/listinfo/zope-web
Re: Re: wiki.zope.org outage [ In reply to ]
> Did you try the .old file left by the pack?

I did; it showed the same error. I agree, I don't think pack caused any
trouble.

I can keep the problem Data.fs around for a little longer, if there's any
other diagnostics that would be interesting to try let me know.

_______________________________________________
Zope-web maillist - Zope-web@zope.org
http://mail.zope.org/mailman/listinfo/zope-web
Re: Re: wiki.zope.org outage [ In reply to ]
On Feb 15, 2007, at 4:30 PM, Simon Michael wrote:

>> Did you try the .old file left by the pack?
>
> I did; it showed the same error. I agree, I don't think pack caused
> any trouble.

Yup. So if you want to reduce the size of the file, you can pack it. :)

> I can keep the problem Data.fs around for a little longer, if
> there's any other diagnostics that would be interesting to try let
> me know.

I don't have time to do anything with it.

You should be suspicious of the disk this database is on.

Jim

--
Jim Fulton mailto:jim@zope.com Python Powered!
CTO (540) 361-1714 http://www.python.org
Zope Corporation http://www.zope.com http://www.zope.org



_______________________________________________
Zope-web maillist - Zope-web@zope.org
http://mail.zope.org/mailman/listinfo/zope-web
Re: Re: wiki.zope.org outage [ In reply to ]
Ok, thanks. Perhaps the problem arose from being too near this VPS's disk
quota.

I pack weekly, I'm going to try moving the wikis into a separate storage to
give me more control.

_______________________________________________
Zope-web maillist - Zope-web@zope.org
http://mail.zope.org/mailman/listinfo/zope-web
wiki.zope.org outage [ In reply to ]
The joyful.com/zwiki.org/wiki.zope.org server was down for 19 hours due to
some insane apache/zope/qmail/curl deadlock that I couldn't figure out.
Service has been restored and plans for better uptime are being hatched.

_______________________________________________
Zope-web maillist - Zope-web@zope.org
http://mail.zope.org/mailman/listinfo/zope-web