larryt@AE.MSSTATE.EDU (Larry Thorne) (03/02/90)
We have a 4D/70GT running IRIX 3.2. All was going fine until 4 days ago when the mount daemon (/usr/etc/rpc.mountd) starting dying. It seems to live until someone tries to mount or umount one of the SGI disks. A possible cause could be that we are mounting the SGI's disks on several more hosts now, and the SGI is in turn also remote mounting disks from all the same hosts (a sort of cross-mounting trick to make all file systems available system-wide). We have a Sun 4/280 & a Sun 4/260 and an Ardent Titan playing this trick with the SGI. Each of the three machines are remote mounting 2 of the SGI's filesystems and the SGI is remote mounting one filesystem from each of these other 3 machines. Also, another client of the 4/280 will occasionally remote mount one of the SGI filesystems, for backup/restore stuff, etc. Also, I'm having to use the -n option on rpc.mountd so that the Ardent isn't locked out of the game. Anyhow, is this the cause of rpc.mountd dying? Is there any way to get an error message from rpc.mountd before it dies? How can I tell what's causing all this? Do I need to start more nfsd processes? Any and all info will be greatly appreciated! Larry Thorne larryt@ae.msstate.edu
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (03/03/90)
In article <9003012042.AA18640@Icarus2.AE.MsState.Edu>, larryt@AE.MSSTATE.EDU (Larry Thorne) writes: > ... > Anyhow, is this the cause of rpc.mountd dying? Is there any way to get > an error message from rpc.mountd before it dies? How can I tell what's > causing all this? Do I need to start more nfsd processes? > > Larry Thorne > larryt@ae.msstate.edu What is meant by "dying"? We run mountd as an inetd deamon, so that it should live only long enough to service one (or more recently a batch) of requests. There have been versions that lived until killed, but I don't think they were shipped. If you are seeing dying as in core files, it would be good to communicate with the Hotline. If you are seeing dying as in inetd or portmap not answering, that is another problem the Hotline should hear about. Nfsd has little to do with mounting. The `rpcinfo -p` command can be useful to see if portmap is running and knows that inetd is willing to start mountd. The `netstat -a` command can be used to see if inetd is in fact listening on the port that portmap thinks it is. Nfsstat can also be useful. Vernon Schryver Silicon Graphics vjs@sgi.com