Outset: 1 NFS server (with lockd) 2 NFS client (with lockd) The clients serve several jails with apache, whose data (www) resides on the server From time to time everything seem to freeze. Then, after one minute or so, the system works again as nothing had happened. In these occasions I get this in the logs on the client madchines: Mar 26 10:29:38 virt1 kernel: nfs server 192.168.40.121:/data/mount_servers/wwwsec/www: lockd not responding followed shortly after by: Mar 26 10:29:38 virt1 kernel: nfs server 192.168.40.121:/data/mount_servers/wwwsec/www: lockd is alive again On the server I only get this: Mar 26 10:29:31 data1 kernel: NLM: failed to contact remote rpcbind, stat = 5, port = 28416 I don't think it's a network problem, since all connections are local and high speed (1Gb/s) I must admit that, with the other nfs problem I reported weeks ago, this kind of freebsd system seems less than stable to me, and this is very disappointing... Anyway I'd appreciate any pointer on this issue...
On Mar 26, 2010, at 3:08 AM, Giulio Ferro wrote:> Outset: > 1 NFS server (with lockd) > 2 NFS client (with lockd) > > The clients serve several jails with apache, whose data (www) resides on the serverIf you need file locking to work reliably, you pretty much have to give up on using NFS + rpc.lockd and run against a local UFS filesystem. Regards, -- -Chuck
On Sat, Mar 27, 2010 at 11:50:17AM -0700, Chuck Swiger wrote:> On Mar 26, 2010, at 3:08 AM, Giulio Ferro wrote: > > Outset: > > 1 NFS server (with lockd) > > 2 NFS client (with lockd) > > > > The clients serve several jails with apache, whose data (www) resides on the server > > If you need file locking to work reliably, you pretty much have to give up on using NFS + rpc.lockd and run against a local UFS filesystem.I thought fcntl(2) worked reliably/properly with NFS on FreeBSD? I remember reading somewhere how mixing locking types (fcntl vs. flock vs. lockf) causes major problems, but as long as the same locking method is used universally things should work...? -- | Jeremy Chadwick jdc@parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB |
On Fri, 26 Mar 2010 12:08:26 +0200, Giulio Ferro <auryn@zirakzigil.org> wrote:> Outset: > 1 NFS server (with lockd) > 2 NFS client (with lockd) > > The clients serve several jails with apache, whose data (www) resides on > the server > > From time to time everything seem to freeze. Then, after one minute or > so, the system > works again as nothing had happened. > > In these occasions I get this in the logs on the client madchines: > Mar 26 10:29:38 virt1 kernel: nfs server > 192.168.40.121:/data/mount_servers/wwwsec/www: lockd not responding > > followed shortly after by: > > Mar 26 10:29:38 virt1 kernel: nfs server > 192.168.40.121:/data/mount_servers/wwwsec/www: lockd is alive again > > > On the server I only get this: > Mar 26 10:29:31 data1 kernel: NLM: failed to contact remote rpcbind, > stat = 5, port = 28416 > > I don't think it's a network problem, since all connections are local > and high speed (1Gb/s) > > I must admit that, with the other nfs problem I reported weeks ago, this > kind of freebsd system seems > less than stable to me, and this is very disappointing... > > Anyway I'd appreciate any pointer on this issue...I'm no NFS expert, so I don't know if I can help, but regularly people want to know your FreeBSD version(s), your used NFS version, if you are running NFS over TCP or UDP. Does it help to switch between UDP/TCP? Ronald.