Strangely, putting a dns for the local zone (and forwarder for the rest) improves the situation. No cannot monitor in kern.log anymore ...
________________________________ De : Bob Proulx <b...@proulx.com> À : debian-user@lists.debian.org Envoyé le : Mardi 20 Septembre 2011 21h46 Objet : Re: nfs problem Stephane Durieux wrote: > I have messages like this in logs: > lockd cannot monitor ip_address > and also messages about statd The lockd: cannot monitor messages are due to the kernel being unable to communicate with the rpc.statd process. That usually indicates that the rpc.statd is not running, possibly due to it crashing. NFS is a stateless protocol. But file locks are inherently stateful. To maintain file locks some auxilary programs are used such as the rpc.statd. Check that the rpc.statd is running. ps -e | grep rpc.statd It is started from the nfs-common service. You may have to restart it. # service nfs-common restart Problems with NFS locking has been a problem forever. > But client ip are present in /etc/hosts. This has nothing to do with host names or DNS resolution. > Do I need to put a DNS server for reverse resolution even for > private ip, it s a pity No. Not for NFS. But generally it is always a good idea that names resolve to addresses and the reverse of resolving addresses to names also works. Bob