Reg NFS problem

[hareesh]

Hi,
We have one 4100 alpha server configured as NFS server with default settings and the os version is 4.0F and we have got around 25 clients with the version of Tru64 5.1, we have configured our clients in such a way that all clients will have one local user called sophos and all clients home directory is same (eg /home_sophos)which is NFS mounted
The probelm what we are facing is ,if i restart a client and try to login to sophos it will take toomuch time to get the prompt in exterminal window and it may affect the other clients also (same xterminal will take time to give prompt)but if i login to root(in clients)and access the nfs mouted directry i will be able to access all files
Since there is no overload on network and NFS server, Iam not able to find the exact reason for this
So kindly help me out to solve this problem

Thanks..
KH.

[Ralf Puchner]

It seems login and access of the shared user directories are slow. Please check used login shells and nfs locking (not all shells uses locking).

Be sure name resolution works fine and is properly configured.

What does the logs tell us? (on NFS server and client side)

[hareesh]

Hi,
Since this problem is happening once in a while only not all times,some times i will be able to login without any problem also.hence there is no problem with the permission and address resolution ,and more over it won't log any errors also.If it related to some file locking (bcos Home directory is common for all clients and due to this all profile and history file will be reading whenever a user logs in)how do i check this is a problem related to file locking or some thread related ..pls advice me..
Thanks..

[Roberto Polli]

Hi!
Did you check your NFS settings? Send here your
/etc/rc.config.common !

Maybe you problem is related to the settings of your NIC. Check the output of
#consvar -g INTERFACE
where INTERFACE is the SRM name of the ethernet card (eg. ewa0, ewb0,...)

Do you use authentication systems like LDAP or NIS? If so, maybe using something like a Name Server Cache daemon could be used to speed up.

A least hint: try to run sys_check, it will return an html doc full of helpful ifo on how to modify your configuration.

Peace, R.

[hareesh]

Hi
iam herewith attached my rc.config file..pls help me out to find the problem


#
DISPLAYTYPE=
HOSTNAME="fs1"
NUM_NETCONFIG="1"
MAX_NETDEVS="24"
NETDEV_0="tu1"
NETDEV_1=
NETDEV_2=
NETDEV_3=
NETDEV_4=
NETDEV_5=
NETDEV_6=
NETDEV_7=
IFCONFIG_0="x.x.x.x netmask 255.255.0.0"
IFCONFIG_1=
IFCONFIG_2=
IFCONFIG_3=
IFCONFIG_4=
IFCONFIG_5=
IFCONFIG_6=
IFCONFIG_7=
RWHOD=
ROUTED=
ROUTED_FLAGS=
ROUTER=
GATED=
GATED_FLAGS=
JOIND=
JOIND_FLAGS=
PAGERAW="1"
PAGEFILE="/dev/rz8b"
PARTITION="rz8b"
PARTITIONTYPE="RZ29B"
PAGEMINSZ=0
PAGEMAXSZ=0
QUOTA_CONFIG=
export DISPLAYTYPE HOSTNAME NUM_NETCONFIG QUOTA_CONFIG
export NETDEV_0 NETDEV_1 NETDEV_2 NETDEV_3 NETDEV_4 NETDEV_5 NETDEV_6 NETDEV_7
export IFCONFIG_0 IFCONFIG_1 IFCONFIG_2 IFCONFIG_3 IFCONFIG_4 IFCONFIG_5
export IFCONFIG_6 IFCONFIG_7 RWHOD ROUTED ROUTED_FLAGS GATED GATED_FLAGS
export ROUTER MAX_NETDEVS JOIND JOIND_FLAGS
export PAGERAW PAGEFILE PARTITION PARTITIONTYPE PAGEMINSZ PAGEMAXSZ
INST_SERVER="rz"
export INST_SERVER
XLOGIN="cde"
export XLOGIN
LAT_SETUP="1"
export LAT_SETUP
ENVMON_CONFIGURED="1"
export ENVMON_CONFIGURED
ENVMON_HIGH_THRESH="40"
export ENVMON_HIGH_THRESH
NFS_CONFIGURED="1"
export NFS_CONFIGURED
NFSSERVING="1"
export NFSSERVING
NONROOTMOUNTS="0"
export NONROOTMOUNTS
NUM_TCPD="8"
export NUM_TCPD
NUM_UDPD="8"
export NUM_UDPD
NUM_NFSIOD="7"
export NUM_NFSIOD
NFSLOCKING="1"
export NFSLOCKING
NUM_PROPLISTD="0"
export NUM_PROPLISTD
PCNFSD="0"
export PCNFSD
AUTOMOUNT="0"
export AUTOMOUNT

[Ralf Puchner]

sorry, but rc.config absolut useless in this situation. Locking depends on mechanism on all involved machines. a variable within rc.config does not mean the startup scripts runs ok and all processes are properly configured!

If you have problems with locking, please go to the troubleshooting section of the network manual! It explains in detail how to check your system.

[Roberto Polli]

Ralph,
can you explain which locking problem can KH encounter?

Shouldn't be a NUMTCP NUMUDP matter? They seems to me quite low.

Maybe running sys_check can help.

What do you think?

[hareesh]

HI,
i have one more doubt that ,since all clients home directory is same (NFS mounted) so whether environment files like .profile ,.sh_history or .dt ,.dtporfile ,etc..will create some problem..??

KH

[Roberto Polli]

Ralf,
sorry to have misspelt your name!

KH,
. shouldn't cause problems if nothing odd is in there. Try to check if there are command you don't know in it and #man for them.

To me it won't matter.

Do you experience some high workload on nfs
eg use
#w
and
#ps
to monitor the status of your nfs daemons when 'slow mounting' happens.

Do you have sys_check on your machine?
Pax, R.

[Ralf Puchner]

Roberto,

e.g. some shells uses locking for their startup files.

Instead of wild guesses and endless discussions start a trace or go through the troubleshooting section of the network manual.

sys_check is a good starter...

[hareesh]

Hi,
Actually i have checked the sys_check output it says that i should increase the NUM_TCPD and NUM_UDP value to 32 (presently it is only 8 and 7 respectvely),The probelm is some time if try with only one client (all other clients i will shutdown)i am facing the same problem so i got confused.so i doubt increasing these parameter will help .

[Ralf Puchner]

haaresh,

the sys_check always suggested this settings, but don't be anger if this will not solve your problem.

sys_check is required to analyze a problem not to solve it by running the tool nor by setting the suggested parameters. So please use the output to verify your configuration.
Troubleshooting guide will give you an starter.

And if you are lost, open a call within a HP support center....

[hareesh]

Hi,
I have observed one more error messages during this problme ,i mentioned it below:

May 6 04:09:01 delta lockd[363]: call_rpc to fs2 STATv1 failed: RPC: Timed out

May 6 04:09:01 delta lockd[363]: rpc.lockd: Cannot contact status monitor!

here delta is the nfs server ,this message i have observed in /var/adm/syslog.dated/current/deamon.log file. But the same time if i give Ctrl+C at the Xterm in clients i will get the prompt and i will be able to work..but if i want to get the prompt immidiatley it will take 15 to 20 mnts..

pls reply if any one get any clue abt this problem
Thanks
KH

[Ralf Puchner]

As written long before, this seems to be a nfs locking problem. If you want to solve it please have a look into the troubleshooting section of the network guide which deals with that problem.

But it seems we are in an endless loop until you read the troubleshooting section.....

[hareesh]

Hi,
Thanks for the sugestions , I got some some solution for this problem,ie , I have to restrt the "rpc.statd" process once again.If i do so the problem is etting resolved immidiatley.so can any one can suggest any other solution or tuning for a parameter,through we can avaoid this ..?
Thanks
KH

[Ralf Puchner]

the status daemon is part of the nfs-locking mechanism controlling the hosts having locks on the machine. It starts if locking is configured during nfs startup.

If statd cores there seems a DNS problem or a invalid locking command. Be sure you are using RFC compliant hostnames, maybe an issue here...

Btw. "man rpc.statd" for more info!

[hareesh]

Hi,
Thanks for ur suggestion actually we don't have any DNS server configured here we hostname resolution is only thru /etc/hosts file . and we r usiong only short hostname not a fully qualified domains,mean while i just wanted to know is there any patch available for to rectify this (rpc.statd) locking problem my NFS server is running on a AS4100 with Tru64 4.0F os..pls help me in this regards

Thanks
HK

[Ralf Puchner]

haresh,

RFC compliance has nothing in common with full vs. short hostnames. Are you sure you understand my answer?

It is a problem on your side not a "bug". So check configuration. If you are lost open a call within the support center because it is too hard to learn you the basics here in the forum.... hopefully you have read the man-page entries of rpc.statd.

A starter is to cleanup the sm directory prior to restarting nfs.