cannot lock/unlock group at ...

[Stas Pirogov]

Hi.

Yes, there is an NFS server involved.
Actually everything except binaries and
configuration is located on NetApp server.

The OS is Solaris 8 with latest patches.

I also noticed that there are additional
messages in error log that I never saw
before:

overchan: malformed token [B6E663D9736337B6E31C5780CE8423E9] alt.binaries.nl:2371315
innd: tradindexed: cannot lock group entry at 65608: Resource temporarily unavailable
innd: tradindexed: cannot unlock group entry at 65608: Resource temporarily unavailable
overchan: malformed token [BA538122EC1A8EDFEE36673F618415B3] control.cancel:14820351
innd: tradindexed: cannot lock group entry at 2392520: Resource temporarily unavailable
innd: tradindexed: cannot unlock group entry at 2392520: Resource temporarily unavailable

It seems that both errors are somehow related, because I
have lock/unlock attempt error for every malformed token.
Sorry that I didn't mentioned that in the first message.

Anything we can do to fix the problem ?

Stas.

On Sun, 18 Jul 2004, Russ Allbery wrote:

> Date: Sun, 18 Jul 2004 09:10:23 -0700
> From: Russ Allbery <rra at stanford.edu>
> To: inn-workers at isc.org
> Subject: Re: cannot lock/unlock group at ...
>
> spam <spam at keyos.org> writes:
>
> > Today after upgrading from INN 2.3.3 to INN 2.4.1 using regular 'make
> > update' I started seeing following error message in general log:
>
> > innd: [ID 801587 local0.error] tradindexed: cannot lock group entry at 2008648: Resource temporarily unavailable
> > innd: [ID 801587 local0.error] tradindexed: cannot lock group entry at 2008648: Resource temporarily unavailable
> > innd: [ID 801587 local0.error] tradindexed: cannot unlock group entry at 2008648: Resource temporarily unavailable
> > innd: [ID 801587 local0.error] tradindexed: cannot unlock group entry at 2008648: Resource temporarily unavailable
>
> > Huge amount of such messages coming in bulks (around 100 messages at
> > once and then pause for a minute or two). The numbers of group entry
> > position in the message change constantly.
>
> > New binaries were compiled with exact same options as old ones and
> > configuration was checked and compared to generated samples.
>
> > I don't see any impact on service, but I would like to understand why
> > there's a problem with locking and to fix it.
>
> Yeah, INN continues on if it can't get a lock.
>
> For some reason, your operating system is occasionally rejecting attempts
> to do advisory locking on the group index file with EAGAIN.  I'm not sure
> why.  It should block the operation until the lock can be acquired, if
> it's currently held by another process.
>
> What operating system is this?  And is there NFS involved?
>
> (INN 2.3.3 wasn't as comprehensive about its locking, so I'm not too
> surprised that this changed between versions.)
>
> --
> Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>
>
>     Please send questions to the list rather than mailing me directly.
>      <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.
>
>