RPC errors (Networker 6.02 / HP-UX Backupserver)

[rruisz]

hi fellows,
we have really big problems with our backupserver(s) we're running networker 6.02 on HP-UX 11.00 (N and L-Class systems).
when a backupgroup starts the nwadmin gui hangs, on the backupserver are a lot of "defunct" prozesses (zombies) - caused by the savegrp-prozess - after about 15min-2 hours the defauncts are gone and "normal" save prozesses are running (also the backup is then running) - nwadmin comes back with normal operation.
during this time its not possible to start further backups (if an addition savegrp is starting it just tells:
save: RPC warning: RPC cannot make a procedure call: Create failed: Remote system error - Connection timed out
save: Cannot open save session with <host>)

our integrator told us - its a network or DNS problem - but now i'm switched to hostfile lookup - and its still the same problem.

does anybody have the same problems - or even better anybody knows a sollution????

pls advice robert

 

[rgzimmer]

Is this happening on all clients or just some? What OS's are the clients?

 

[rruisz]

it's happening an all clients (we have 2 backupserver - 1 for the unix (client) area and one for the nt (client) area - the clients OS are: hp-ux 10.20 and 11.00, solaris 2.6 and 8, win nt4 and w2k, oracle 8.1.7 (not really an OS )but not all the time - so once it happens in a group of HP-ux systems, an other time it happens when we run a backupgroup of mixed OS clients - it also doesnt happens during the same time of the day......

 

[rgzimmer]

I would lean towrd a network problem since it is intermittent. If it was DNS related, you would see the problem all of the time.

Check to see if the servers and clients are using 100mb full duplex. That may help some. Without knowing much about your acrhitecture, it's hard to hazard a guess as to what it may be.

 

[Zantor]

I've noticed that the RPC errors occur whenever the Networker Server can't connect to the Client. This can be because the client is down, the network is slow, reverse lookup isn't working, the RPC service is down on the client, the client nsrexecd.exe process is hung, etc. Kind of a general purpose error message.
What OS is your Networker Server running. If it is NT we have found that 150 clients is about the maximum. If you offload some of the work to a storage node you may find a lot of your problems mysteriously disappear.
We have found that we can easily swamp the networker process on the server even though the server itself is always underutilised.
You could also try reducing the number of allowable concurrent sessions. We found that this also helped.

Good luck

 

[piotrwalczak]

Hi,
I had (have?) similar problems running NW 5.5.4, then 6.1 on HP-UX 11.0. It seems that the problem is solved by setting the network cards on the server and on the clients to NOT autonegotiate. Hope this helps.
Good luck.
Piotr