CDE Login is slow for NIS Users

[V.Thirumalaimurugan]

Whenever root user logs in it opens up the desktop quickly.At the same time when a NIS User logs in it takes 6-7 mins to show the desktop.
Any suggestion on this.

[Robert Gamble]

I've seen this before if the user's home directory is actually NFS mounted from another system and the LAN is having issues. Network issues can cause this (wrong port negotiation settings, packet storms, etc).

Try issuing a 'bdf' to see if it is hanging, Vs. a 'bdf -l'.

Hope this helps!

[Stuart Abramson_2]

We used to have this problem, years ago.

If user home directories are NFS mounted from a NFS home directory server, and if the HISTFILE is located in the home directory, which it is by default, then login is delayed by NFS activiy , LAN, network traffic, etc.. The login process wants to write to HISTFILE, but can't get it because of NFS/LAN delays.

So you have two choices:
1. Improve NFS performance by messing with the number of "biod" and "nfsd" daemons on the server and clients, and other techniques.

2. Move the HISTFILE to a local directory:

vi all .profiles (maybe /etc/profile)
export HISTFILE=/tmp/.sh_history.$LOGNAME.$$

Then write a cleanup script in "root cron" to clean out old HISTFILEs in /tmp.

I would do both!

[Tim Sanko]

Again I point the hairy finger at DNS/NFS. We have seen long logins from automounted nfs drives with DNS being overloaded/slow.

Tim

[Shannon Petry]

This is not just from NFS, but more commonly name resolution. Try to add host entries locally. If this fixes the problem, make sure that /etc/nsswitch.conf is configured correctly.

If your not familiar with editing this file, use SAM -> Networking and Communications -> nsswitch

Regards,
Shannon

[Garry Ferguson]

Even with NFS slowness due to possible network
problems I think 6/7 minutes is VERY long.
You could try adding some debugging "echo"
statements in a user's "login" or "profile"
file to narrow it down.

We have found that some logins try to tell the
user his quota. If he has quota set on remote
machines this can be slow.

Have you removed another machine recently?

Regards, Garry Ferguson

[Thirumalaimurugan.V]

Robert -- there is no issue in the bdf output.It is not hanging.

I tried by creating local history file but still the problem presits.

I got the similar kinf of systems having similar NFS home directory from the same filer.Those systems are working fine only on this i could see the problem.

I observed the slowness after installing the Sep 2002 QPK1100.

[Robert-Jan Goossens]

Hi,

check this patch, there are a lot of bugfixes for nis in combination with nfs.

http://www5.itrc.hp.com/service/patch/patchDetail.do?patchid=PHNE_28567&context=hpux:800:11:00

Hope it helps,

Robert-Jan.

[Alex Glennie]

here's a thought .... you say it started occurring after a patch install ....

swlist -l fileset | grep -i CDE

swremove the latest CDE patch you have installed .... does the problem persist ?

I'm wondering if the functionality of dtdbcache has changed in some way.

what's in /etc/inetd.conf
Next lets get an idea of where the hang occurs :

what the last thing you see on the screen prior to the hang ?

Does a login to a CDE failsafe session encounter similar problems ?

[V.Thirumalaimurugan]

Alex i didn't find any difference in the swlist -l fileset |grep -i CDE from the problematic system with another working one.

Failsafe mode works fine without any above said problem.