ack@caldwr.caldwr.gov (David Ackerman) (01/14/88)
How many sites out there are using NFS? We have a Sun mounted on our
network root so it look like just another node. The Apollo network root
is mounted as a single file system on the Sun so the Apollo ring
looks like one big file system. So far so good.
But all is not wine and roses. Since an initial period with no problems
after bringing up NFS, we have begun to have problems with our
tcp server dying. It won't die for any particular reason, such as being
overloaded by the nfsd. Sometimes it will die when there's nothing
noticeable going on, and sometimes it will die when you try to cd into
the NFS partition.
I have also noticed that the CPU usage for nfsd is quite high, much
higher than the tcp server itself. Why didn't Apollo allow for multiple
nfsd's the way some other companies (who shall remain nameless) do? It
would take the load off of the lone nfsd.
If anyone has seen the above problems (or had different ones with NFS)
and has overcome them, please send me e-mail. AdvTHANKSance...
David Ackerman
California Department of Water Resources caldwr!ack@ucdavis.edu (Internet)
"It's the water, and a lot more..." ...!ucbvax!ucdavis!caldwr!ack (UUCP)
The opinions expressed above are mine, not those of the State
of California or the California Department of Water Resources.giebelhaus@hi-csc.UUCP (Timothy R. Giebelhaus) (01/20/88)
In article <280@caldwr.caldwr.gov> ack@caldwr.caldwr.gov (David Ackerman) writes: >But all is not wine and roses. Since an initial period with no problems >after bringing up NFS, we have begun to have problems with our >tcp server dying. > [...] >Why didn't Apollo allow for multiple >nfsd's the way some other companies (who shall remain nameless) do? It >would take the load off of the lone nfsd. I had some TCP problems, but since I installed the patch to TCP 3.0 life has been wine and roses. After installing the patch, I run routed with the -r arguement to tcp_server instead of from a line in my etc.rc file. I am not running 9.7 yet because I don't have TCP 3.1 yet. I have not been running nfsd much either since I have been waiting for the other unix boxes to standardize on some set of uid's (almost done). But, maybe news of the 3.0 patch may help anyway. Apollo does allow for multiple nfsd's. You can run it on as many nodes as you like (presuming you have the license to do so). In the Apollo NFS manual (the best NFS manual by far which I have seen) it gives several configuration suggestions including running multiple nfsd's for higher load networks. -- --------------------------------- UUCP: {uunet, ihnp4!umn-cs}!hi-csc!giebelhaus ARPA: hi-csc!giebelhaus@umn-cs.arpa Nobody I know admits to sharing my opinions. I don't even have a pet.