moorej@dvinci.usask.ca (Jonathan Moore) (12/07/89)
Hello. I have a few questions which are all loosely related, so I will put them all in one posting. First, some background. I am running a Sun 3/60 as a file server for a medium-sized PC network (~40 MS-DOS machines) in a microcomputer lab. The PC's are running PC-NFS 3.01. The 3/60 is running SunOS 4.0.3. I have various home-grown RPC programs running on the server which seem to run OK. Now for the questions: 1) Every so often (about once a week or so), clients stop being able to mount the server's disks. Terminal sessions are still OK. This affects the PC-NFS clients as well as other UNIX servers. Nothing I try short of rebooting the server can fix the problem. Is there something less drastic that can be done about this? Is something perhaps conflicting with the NFS stuff? 2) The /etc/rmtab file on the server seems to grow (and grow...). Many of the entries have a hash character in them. The file temds to look something like: . . #kul13:/dos #kul18:/dos skul21:/dos . . I assume PC-NFS is putting the "#" hash characters in the lines when a PC client unmounts (or maybe when it dies). Is this something to cause concern? Can I just periodically trash this file (It gets to be >1 Mb)? 3) Another PC-NFS question. Is there any way to (somewhat accurately) account for usage under PC-NFS? It would be advantageous to be able to track how much the lab is being used and by whom. However, it seems that the stateless nature of NFS makes this type of accounting impossible, since the server doesn't keep track of what the clients are doing. I have tried running a simple accounting system that starts up when the PC is booted, and stops when the user logs out (there is a "logout" program that, among other things, tells the accounting program the user is done). However, this has problems, since the users often turn off the machine prematurely, or it crashes altogether. Any insight anyone has into NFS-type accounting would be greatly appreciated. I would appreciate any comments/answers/etc regarding any of these questions that anyone may have. Thanks very much in advance, Jonathan. -- |--- Jonathan Moore ------------------|-- Voice: (306) 966-5295 ---------| | College of Engineering | | | University of Saskatchewan | Email: moorej@sask.usask.CA | |___ Saskatoon, Canada S7N 0W0________|__________ moorej@dvinci.usask.CA _|
geoff@hinode.East.Sun.COM (Geoff Arnold @ Sun BOS - R.H. coast near the top) (12/07/89)
In article <1989Dec6.163953.20378@dvinci.usask.ca> moorej@dvinci.usask.ca (Jonathan Moore) writes: > 1) Every so often (about once a week or so), clients stop being > able to mount the server's disks. Terminal sessions are still OK. > This affects the PC-NFS clients as well as other UNIX servers. > Nothing I try short of rebooting the server can fix the problem. > Is there something less drastic that can be done about this? Is > something perhaps conflicting with the NFS stuff? There's not enough information to go on here. Is the mount daemon still running? (Try "showmount -e hostname".) Are the nfsd daemons hosed? (Check with ps ax.) Is it only mounting that fails, or are existing mounts screwed up? Are there any anomalous stats shown by "nfsstat -s"? Etc. Etc. > 2) The /etc/rmtab file on the server seems to grow (and grow...). > Many of the entries have a hash character in them. > . > > I assume PC-NFS is putting the "#" hash characters in the lines > when a PC client unmounts (or maybe when it dies). Is this > something to cause concern? Can I just periodically trash this > file (It gets to be >1 Mb)? Way back when (when? maybe SunOS 3.x.) I thought that "rpc.mountd" tried to actually delete entries from /etc/rmtab, but maybe I was dreaming... Anyway, the SunOS 4.0 code includes the fragment: void rmtab_delete(pos) long pos; { if (f != NULL && pos != -1 && fseek(f, pos, 0) == 0) { fprintf(f, "#"); fflush(f); } } which is responsible for the '#' lines. I suggest that you just trash the file if it gets too big - it's purely advisory, anyway. BTW, PC-NFS isn't writing anything - it's the rpc.mountd on the server. > > 3) Another PC-NFS question. Is there any way to (somewhat > accurately) account for usage under PC-NFS? It would be advantageous > to be able to track how much the lab is being used and by whom. The only straightforward solution to this seems to be to write a little DOS TSR that is kicked off when you log in and which opens an NFS-mounted file and periodically stuffs a timestamp in there. This would run until logout or (obviously) power-off. A Unix program can then browse these files and compute the usage. The TSR would look a little like the print redirector in structure. Taking advantage of NFS seems the best way to keep the size down. Geoff Arnold, PCDS Group, | Quote of the week: "Shut up and mind your Sun Microsystems Inc. | Canadian business, you meddlesome foreigner." Internet: geoff@East.Sun.COM | (Theodore A.Kaldis, <kaldis@topaz.rutgers.edu> Disclaimer: Obviously.... | on Nov. 22 1989, in reply to Joe Nunes.)