If you chose to install the NLM (Network Lock Manager) package when you installed NFS, lockd and statd will start automatically each time the host boots. If you elected not to install the NLM package, you will have to do so using the NFS installation floppies.
If a user requests a lock that already exists, the request is
blocked (refused). Most locks exist only briefly.
block_timeout is the interval that lockd will wait
before retransmitting a blocked request to the server lockd.
By default, lockd will wait 5 seconds before retransmitting
the lock request. You can alter this by setting the -b flag
when you invoke lockd on the client:
/etc/lockd -b 30
This will cause the local lockd to wait 30 seconds before retransmitting a blocked lock request to the server. It is generally desirable to change the block_timeout when you wish to minimize network traffic or if you are transmitting to a slow server.
After a host system has crashed and come back up, there is
a brief period during which no new lock requests are
processed. The only lock requests accepted are reclaim
requests from clients that already have outstanding locks.
By default, this grace period is 40 seconds. You can alter this by setting
the -g flag when you invoke lockd:
/etc/lockd -g 120
This will cause the local lockd to wait for 120 seconds after crash recovery before it processes new lock requests. It is desirable to change the grace period if there are many clients attempting to reclaim locks on files.
By default, lockd allocates two kernel-based server
processes, one for TCP and one for UDP, that handle
lock requests from clients. You may allocate any number of lockd
daemons for either transport protocol by using the -t and
-u flags when you invoke lockd on the server host:
/etc/lockd -t 3 -u 3
This will allocate three lockd daemons for lock requests
transmitted via TCP and three daemons for those transmitted via
This will allocate three lockd daemons for handling UDP lock requests. This is desirable when you wish to process more lock requests or allow the lock requests to be processed faster.