NFS not responding+panic on full quota'd ZFS

[Dennis Glatting]

On Wed, 2012-10-10 at 10:30 +0200, Robert Schulze wrote:
> Hi,
> 
> we are running several webservers on 8.1-RELEASE-p12 which deliver their 
> data from NFS. Ever since we began using ZFS on the fileserver, we 
> always had to track quota usage, because of the annoying "nfs server not 
> responding" error: when a exported filesystem was full and got write 
> access (including unlink), the server just stopped serving.
> 
> As of this night, things became more complicated. Webservers either 
> crashed in sync with a "sleepable thread owns a non-sleepable lock" 
> panic or after the following:
> 
> nfs server 10.0.0.XX:/home/www/XX/YY: not responding
> nfs_getpages: error -484759920
> vm_fault: pager read error, pid 9467 (php-cgi)
> 
> The not responding filesystem was nearly out of quota.
> 
> Does anybody know, if these issues with quota'd nfs-exported zfs are 
> fixed in 8.3?
> 


Similar problem, maybe. Does the network still work (ping, etc)? Can you
login and run commands? Are you running top in another window? If so,
what does it say?

I have four ZFS servers (similar, not exact), NFS clients (little
activity), that hang under heavy local disk and computational load where
the processes are stuck in tx->tx (al la top). A reboot is necessary. 

I gathered equipment for more round of trying to extract debug
information but it is looking like I am going to have to strip ZFS, if
not FreeBSD, from a production environment.



> with kind regards,
> Robert Schulze
> _______________________________________________
> freebsd-fs at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-fs
> To unsubscribe, send any mail to "freebsd-fs-unsubscribe at freebsd.org"