SUMMARY: rpc.nisd problem

From: Shawn Kondel (shawnk@sunfs.math.usu.edu)
Date: Tue May 11 1999 - 12:52:53 CDT


Problem:

Ultra Enterprise 450
sunfs# uname -a
SunOS sunfs 5.6 Generic_105181-06 sun4u sparc SUNW,Ultra-4

Someone sent a 300 meg postscript file into the print spooler. That filled up
the /var slice (also where the nis+ database reside) causing the system to
grind. I removed the ps file and saw that nis+ still working; however, I cannot
view the the nis+ tables using niscat. I have noticed the 80% swap space is
used up, trans.log has gotten huge for sometime, and some of the recently
created user accounts keep getting purge from the nis+ table after a period of
time.

I did some reading and try to do nisping -C (checkpoint) to trauncate the nis+
log files, it failed

sunfs# niscat passwd.org_dir
can't list table: Server busy, try again.
sunfs# tail /var/adm/messages
Apr 29 11:07:05 sunfs nisd[207]: rpc.nisd: unable to fork readonly listener
process.
Apr 29 11:13:44 sunfs last message repeated 1157 times
Apr 29 11:13:45 sunfs nisd[207]: rpc.nisd: unable to fork readonly listener
process.
Apr 29 11:20:25 sunfs last message repeated 685 times
Apr 29 11:20:25 sunfs nisd[207]: rpc.nisd: unable to fork readonly listener
process.
Apr 29 11:27:01 sunfs last message repeated 901 times
Apr 29 11:27:05 sunfs nisd[207]: rpc.nisd: unable to fork readonly listener
process.
Apr 29 11:27:46 sunfs last message repeated 62 times
Apr 29 11:27:46 sunfs nisd[207]: Couldn't fork a process!
Apr 29 11:27:46 sunfs nisd[207]: rpc.nisd: unable to fork readonly listener
process.

Last two created by niscat and the rest are from nisping -C

I did a ps -el and noticed that rpc.nisd has gotten really large.
(I came to a conclusion that the rpc.nisd has gotten so large due to the large
trans.log file and rpc.nisd may used up most of the swap space.)
 
I checked the nis+ troubleshoot documentation and the message "Server busy" is
caused by either not enough swap space or not enough processes.

I enlarged the swap space by adding a swapfile (now have 98% free space) and
freed up most of the processes. But I still cannot view tables or do checkpoint.
Rpc.nisd`s process cannot be stop or nis+ commands will not work without it.

Rebooting the system did not help either.

Solution:
By experiment, I did a nisbackup to a new location
Restored the old nis+ tables (about six month old)
I can view the tables, but the other nis+ clients are having hovac since it
could not authenicate the nis+ server. This nis+ master have cred passwd changed
several months back. (Another story)

I restored the recent backup nis+ tables
It works.

Both rpc.nisd's process and trans.log size has been greatly reduced.
(I thinks the trans.log gets purged when you do a nisrestore)

Moral to the story, do nisbackup and nis+ checkpoint often.
The documentation suggest that you stick "nisping -C" in cron.

I like to thanks for those who replied; eventhough, it didn't solve it.

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Shawn Kondel Computer Specialist
Department of Mathematics & Statistics Email: shawnk@math.usu.edu
Utah State University Phone: (435) 797-4061
3900 Old Main Hill Fax: (435) 797-1822
Logan, UT 84322-3900 Web: www.math.usu.edu



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:19 CDT