SUMMARY : NFS locks (with a NETAPP filer)

Ok,

thanks again for everyone's help:-

Alexandre Perematko [alexp@novator.com]
David Foster [foster@dim.ucsd.edu]
Paul Updike [updike@msai.mea.com]
Mark Hollow [mhollow@commercenti.com]
Richard Smith [richard.c.smith@gte.net]

The biggest suggestion was the LockFile directive in Apache's httpd.conf
being the problem.
This lock file should not be placed on an NFS mounted volume (our ServerRoot
was).
I initially ignored this, as it refers to the USE_FCNTL_SERIALIZED_ACCEPT
and USE_FLOCK_SERIALIZED_ACCEPT compile options which I didn't think we
used. It turns out that we have SE_FCNTL_SERIALIZED_ACCEPT on by default
(httpd -V will tell you what compile options httpd was built with)
If you don't hear from me again, then this has fixed the problem!

Thanks once again


Jez


> -----Original Message-----
> From: Jez Ahl [mailto:Jez@andelconsulting.com]
> Sent: 11 July 2001 12:22
> To: Sunmanagers (E-mail)
> Subject: NFS locks (with a NETAPP filer)
> 
> 
> 
> Hi managers,
> 
> I wonder if any of you Netapp users have seen this problem? 
> (Or any other
> suggestions welcome!)
> 
> We have a couple of Solaris 7 e420s, running Oracle and 
> Apache off an NFS
> mounted (Netapp F740) volume.
> Occasionally (about once a day) one of the 420's will grind to a halt.
> Investigation shows that Apache is spawning lots of extra 
> processes (as it
> says it will do if it doesn't have 5 idle servers). At the 
> same time I get a
> message on the filer saying "Lock manager version 4 (for UDP) 
> not registered
> at host 192.168.105.2".
> It looks like the apache processes are locking on a NFS file. 
> This causes
> all apache processes to become busy, so more are spawned.
> After about 10 mins, the box recovers.
> 
> Is there any way to find out what is being locked on the NFS mount?
> 
> Thanks in advance (again!)
> 
> Jez
> _______________________________________________
> sunmanagers mailing list
> sunmanagers@sunmanagers.org
> http://www.sunmanagers.org/mailman/listinfo/sunmanagers
>