> Hi everyone .
> I m running a Solaris NFS server along with LINUX NIS server .
> On the client side I ve all Linux machines (Fedora 2 n RH 7.2)
> Normally things work fine but at times NFS server gets stuck n
> following error msgs are displayed ->
> nfs:server 10.6.5.27 OK
> nfs:server 10.6.5.27 not responding, still trying
> nfs:task 10725 cant get a request slot
> n all processes trying to access user's nfs mounted files go in to
> D(Disk I/O) status ..n I cant even kill them coz D status processes
> are not in process queue the only option remains is either wait for
> I/O or reboot comp.
> any body faced the same problem ..pls put some fundaes.
> -- Vivek
Depending on the number of LINUX clients, you may need to increase
the thread count on the Solaris NFS server. I have seen this type
of problem before at client sites where the default NFSD count was used.
The default is 1, which is VERY low.
You can set these by (from www.sun.com)
Solaris 8 and prior
The script /etc/init.d/nfs.server sets the maximum number of nfsd
server threads that can run at nce.
For Solaris 9 and subsequent releases
The maximum number of nfsd server threads is configured using the
parameter NFSD_SERVERS in the /etc/default/nfs file.Those threads are
created and destroyed dynamically. They each
consume 16K of kernel stack and most likely handle one NFS filesystem
block. They also run at a higher priority than the timeshare class.
Since the memory requirement is cheap and dynamic, there are no big
drawbacks to setting this value much higher on an NFS server
(a conservative nfsd thread value = num_of_cpus*128).
For Solaris 8, edit /etc/init.d/nfs.server :
/usr/lib/nfs/nfsd -a 1024
For Solaris 9, edit /etc/default/nfs:
Then kill and restart the daemon
NFS server threads run at a high priority. With a really high number
of threads they may potentially starve other non-NFS related work. In
this case, Solaris 9 - Solaris Resource Manager (SRM) may be used
to limit the CPU time allotted to NFS.