NFS Performance Tuning for HP-UX 11.0 and 11i Systems

nfs performance tuning for hp-ux 11.0 and 11i systems page 62
Notes:
Page 62July 22, 2002
Copyright 2002 Hewlett- Packard Company
rpc.lockd
&
rpc.statd
How are NFS file locks recovered
after a client or server reboot?
text text
CLIENT
17
2
3 4
15
1 16
SERVER
11
8 9
10
14 7
6
512
13
rpc.statd
rpc.lockd
kernel
application
rpcbind
rpc.statd
rpc.lockd
kernel
application
rpcbind
17. Notification process takes place between client’s and server’s rpc.statd
daemons during a “change in state”
A state change refers to any time where lockd and statd are stopped and restarted,
either due to the client or server rebooting, or when the daemons are killed and
restarted manually. In this situation, the statd daemon is instrumental in notifying
the remote system, either client or server, of the change in state on the local system.
If the NFS client is the system experiencing the state change then its statd notifies
the statd process on every NFS servers that the client was performing file locking
with, letting those systems know they should discard any locks they are currently
holding for the client.
If the server experiences the change in state, any locks it was holding prior to the
state change are gone. The server’s statd therefore needs to contact the statd
process on all NFS clients who have issued file lock requests with the server,
informing them that any locks they were holding with the server are gone and must
be reclaimed within 50 seconds or the server could give the locks to another client.