NFS - RPC Problem

[Ionut Grigorescu_2]

Hi,

I'm back with an unsolved problem. I have 2 NFS HPUX11.00 server/client and 3 clients(also HPUX11.00). When I was trying to copy a file in a mounted directory (/m/home) I had very bad performance, for example 20 minutes (!!) for 100KB file. I have sniffed with Ethereal and I have seen that some RPC request were responded after a long time or not at all.
I have opened a helpdesk request to HP and I've been told to install some PHKL patches (in their readme files I didn't see anything related to NFS)
After that, I have tried again to copy a file (/stand/vmunix to /m/home/ionut) being root. I have received a RPC write timeout (in syslog.log). When I have tried to copy a file from the same directory (/m/home/file to /m/home/file.old, where file is about 1,9M) cp hangs for a long time (about 45 minutes) then the file is copied but without permissions (something like --------- instead of -rw-rw-r--, like the file wasn't closed).
Also it takes a loooong time to login as , where user needs a mounted home directory, but bdf shows me after that all the exported directories from the server..
I'm using NIS not DNS.
Can you give me a step-by-step troubleshooting guide ? (I have checked a lot of things, but maybe I have missed something

[Robert-Jan Goossens]

Hi,

Fact: you are using NFS and NIS.

Question: Do you use the automounter to mount the homedirs ?

Question: have you tryed ruling out any network problems ?

server1
# time rcp -p /stand/vmunix.prev server2:/tmp/

server2
# time rcp -p /tmp/vmunix.prev server1:/tmp/

Kind regards,

Robert-Jan

[Ionut Grigorescu_2]

Yes, I'm using automounter. I have tried also rcp and I receive a timeout, too.

[Ionut Grigorescu_2]

I have send to HP also the outputs from lanadmin and show conf from the Cisco Catalysts interfaces - there is no problem concerning the interfaces speeds and full/half duplex...

[Ionut Grigorescu_2]

Another thing: MC/Service Guard is running on the servers and comm package is responsible for exporting some directories - /d/home, too. There is a script under /etc/cmcluster - sgscusmx.cf containing some exportfs lines and some options for export.

[Robert-Jan Goossens]

Hi,

starting a search on your problem (automount + nis) will let yoy know results as soon as possible.

Kind regards,

Robert-Jan.

[Robert-Jan Goossens]

Hi Ionut,

Could you try mounting a nfs mount point on your client without using the automounter.


http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B1031-90048/B1031-90048_top.html&con=/hpux/onlinedocs/B1031-90048/00/00/5-con.html&toc=/hpux/onlinedocs/B1031-90048/00/00/5-toc.html&searchterms=nis%7crpc%7cnfs&queryid=20021113-062453

Robert-Jan.

[Ionut Grigorescu_2]

The problem is now solved - after long time...
The problem was there is another guy responsible for CISCO Catalyst and network infrastructure - cabling, switches, routers, etc.. I have seen som inbound errors and discards with lanadmin and I have told him in the Catalyst if there is a missmatch regarding speed, full/half duplex and auto-negotiation. He gave me a print-out from show port and everything was OK there. I decided then to use another Catalyst port and then check the cable. The Catalyst guy discovered that his table with the correspondence between IP addresses and Catalyst ports (was an Excel table) was wrong! We have all the time looked at the wrong ports. The mistake was that the port had auto-negotiation set to ON and I have on my servers set to OFF. Now is everything OK, thanks anyway for support!