SUMMARY: 2 min. ping

From: Paul A. Hepp (paul@naicu.edu)
Date: Sat Oct 30 1993 - 03:48:24 CDT


Thanks to all.

The orignal problem is attached at the end of this message.
The short version of problem: 2 min. ping (server to client)
2 min. passwd resolve, ypcat hosts returned nothing, df worked great
and ps also. Ping from client to server, and other activity was normal.
This was a NIS problem.
ADDITIONAL SYMPTOM: ping `IP address` worked normally.

The solution was to kill and re-start ypbind and ypserv.
>From: Elmar Kurgpold <ekurgpol@Law.USC.EDU>
      jayl@lattice.com (Jay Lessert)
Special thanks to Michael.
      canuck@masc38.rice.edu (Mike Pearlman)
      kill -9 "ypbind process id"; kill -9 "ypserver id";
      /usr/etc/ypserv; /usr/etc/ypbind

>>Additional tips and tests from others.

>From: leclerc@austin.nam.slb.com
  .
  .
  -MOVE TO 4.1.3 as many bugs are removed
>>SEE all patches needed for 4.1.3 at mid-list

>From: perryh@pluto.rain.com (Perry Hutchison)
>From frank@fotlan.army.mil Thu Oct 28 07:19:02 1993
>From: bert@penril.com (Bert Robbins)
>From: mharris@jpmorgan.com (Michael Harris - Portfolio Analysis)
>From: canuck@masc38.rice.edu (Mike Pearlman)
>From afm@ufnmr1.health.ufl.edu Thu Oct 28 12:08:02 1993
>From: Elmar Kurgpold <ekurgpol@Law.USC.EDU>
>From: birger@vest.sdata.no (Birger A. Wathne)
  You should have added some info from programs like
  'vmstat 5' 'vmstat -S 5' 'iostat 5', etc....
  Also a 'ps -agxvww' could be useful.
  From: birger@vest.sdata.no (Birger A. Wathne)

  Does ypcat / ypmatch take a long time? Is the result from
  ypmatch <hostname> hosts' correct?

  Try running 'pstat -s' to see how much swap you have left (your free
  list was rather small), and try running 'vmstat -S 5' over a longer
  period of time. You have 2 xnews processes running, one on the console,
  and one for NeWSprint. You could try restarting the NeWSprint one (I
  think you can reset it by printing one of the documents in the
  NeWSprint direcory, or by simply killing it). It seems to eat a lot of
  memory.
  .
  .
  Try looking at the 'so' and 'po' numbers in 'vmstat -S 5' while the host
  is slow. They should tell you if you have too little physical memory.
>From: jayl@lattice.com (Jay Lessert)
  Try "trace ping anyone", see where the stop is happening.
   Lots of NFS mounts? Possible quota-checking problem. Try
  the "ln -s /bin/true quota" trick.
  Vmstat show any paging?
  Any console messages?
  Using NIS? Problem w/NIS server? Try killing ypbind &
  restarting with -ypsetme, move around to different NIS server.
  try doing w/o NIS completely.
>From rwolf@dretor.dciem.dnd.ca Thu Oct 28 15:18:51 1993
  You are probably got too much network traffic by the sounds of it. Have
  you considered upgrading to SunOS 4.1.3?
  If you do you should add the following patches afterwards:
    Patches installed on accord
    ===========================

    machine: sun 4/40
    arch: sun4c

    100075-11 rpc.lockd
    100103-11 file permission security
    100170-10 jumbo ld
    100173-10 jumbo nfs
    100185-01 /etc/rc.local
    100224-06 jumbo mail rmail
    100249-07 jumbo automounter
    100272-06 in.comsat
    100283-03 in.routed
    100296-04 rpc.mountd
    100305-11 lpr lprm lpc lpstat
    100359-06 jumbo streams
    100372-02 tfs & c2
    100377-05 sendmail sendmail.mx
    100383-06 rdist
    100407-07 accounting
    100421-03 jumbo rpc.rexd
    100448-01 not required since nfs mounted
    100452-42 not required since nfs mounted
    100462-25 not required since nfs mounted
    100465-02 dns res_send resolv.conf
    100482-04 ypserv ypxfrd dns
    100493-02 not required since nfs mounted
    100507-04 jumbo tmpfs
    100512-03 not required since nfs mounted
    100513-02 jumbo tty
    100524-06 not required since nfs mounted
    100567-04 icmp redirect security
    100573-03 not required since nfs mounted
    100584-03 tcp loopback
    100623-03 jumbo ufs
    100650-02 ipcs
    100891-04 libc
    101080-01 expreserve security

>From: tim@ben.dciem.dnd.ca
  You could try watching the IP traffic (etherfind -ip -host IPC_host)
  and then try the ping from the IPC. See just what the IPC is doing
  during that 2 minute period.
>From: glenn@uniq.com.au (Glenn Satchell - Uniq Professional Services)

   I hope that you've got plenty of memory in th eserver as 22 diskless
 clients is a huge load. Since the cpu is so idle it sounds like the
  system is paging heavily or swapping processes out, including your
  login shell. I thiunk that it is time you bought some disks for at
  least a few of those diskless clients and took the load off the
  server.

  Try using "vmstat -S 5" to have a look at the swapping activity. I
  think that you will see some non-zero numbers in the so (swap out)
  column.

  Other ideas are to run top and see what heavy processes are running.
   One of those nfs clients may be beating your nfsd's to death.

***************The Orignal Message. ***********************
Vitals: SS2, 4.1.1, OW3, Read the FAQ, Read _Managing NFS and NIS_
        SS2 is a server for 22 diskless clients and 1 IPC

System is up and users are complaining about slowness.

Server symptoms: login takes ~2 minuntes.
                 ping anyone takes ~2 minutes
                 ping -s anyone; 2minutes to start
                        but avg 1ms respones when running
                 df command; instant response
                 vmstat shows CPU idle 87-95%

Recent "changes": tried to add another diskless client, and
                got variuos problems, so I backed out of that
                ordeal thinking that was what is causing the
                "slowness". No help.

This has got to be a simple problem that I can not see.
Rebooting is not an option today.

Thanks, I will summarize.

**************************************THE END



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:08:27 CDT