For a while I've been seeing errors of this sort in my /var/log/messages kernel: nfsd: too many open TCP sockets, consider increasing the number of nfsd threads
I googled a bunch and think the solution might be to boost RPCNFSDCOUNT in the line "[ -z "$RPCNFSDCOUNT" ] && RPCNFSDCOUNT=8" in the file /etc/init.d/nfs. Question: This seems a suspicious place to change it. Isn't there a nfs config file somewhere else? Question2: How high can I boost the number of NFS threads? Or how I should I? Is there any metric I can track to decide an optimum number? Most recomendations were for 32, 64 or 128. What do people suggest? Any downsides to having numbers that are too high? In case it matters, this is our master-node Linux server and has the /home directories for each user exported to about 200 odd compute-nodes. -- Rahul _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
