A solution was suggested to use NFS v4, as it builds lease-based file locking directly into the protocol itself, dropping the need for rpc.lockd.
Since OpenBSD does not implement NFS v4, I am now using the following client side option: soft,timeo=30,retrans=3. Will see how it works. On Friday, May 15th, 2026 at 10:37 AM, Otto Cooper <[email protected]> wrote: > > /var/log/daemon > ... > > May 15 09:51:51 openbsd rpc.lockd: duplicate lock from <clientid>.26203 > > May 15 09:51:51 openbsd rpc.lockd: no matching entry for <clientid> > > Hello, > > The above log entry occurs frequently when a NFS client is connected via > Wi-Fi. > The client has its MAC on record, and a static IP assigned by DHCP. > > I see the problem with MacOS clients. > I do not have evidence of the problem with other OS, not because they are > immune but because I have not tested them yet. > > I suspect that when the client disconnects, because of sudden loss of signal, > the client does not unlock the resource. When it reconnects, client and > server are misaligned: the server holds the lock and expects the client to > use it, but the client issues a new lock request instead. Or perhaps > something else occurs. > > As this problem is occurring frequently, it is causing pains. > > Is anybody else seeing this? How did you solve the problem? > > Thank you. > > -- > Otto
