Ladies and Gentlemen:
The first response told me what to do to find the problem, which was as
he guessed, some sort of loop. I didn't know enough to figure out the
next question, which was WHY? Coincidentally, I had planned to re-boot
the server to add some things to the automount maps and the re-boot
of course cleared up the looping problem.
As to why run the automounter on the server rather than static mounts,
many of the workstations export data on their local disks to the whole
network.
The original question and the responses are at the end of the message.
Thanks to all,
----------------------------------------------------------------
| Hap Hinrichs G W Hinrichs, III |
| Research Director |
| gwh@ecliptic.stat.nielsen.com A. C. Nielsen Co. |
| voice: (708) 498-6320 x2430 Nielsen Plaza |
| fax: (708) 205-4014 Northbrook, IL 60062 |
----------------------------------------------------------------
>>> Original message and responses
I had top fired up on the main server (4/490, 4.1.1, 128Mb) to check
on a couple of tape jobs and noticed that the automounter was right
in there gulping up CPU etc with the user jobs. I don't think my
memory is that far gone yet, but I don't remember that being a normal
state of affairs. Is this normal? If not, what can I check to try
to isolate the problem?
Output from top:
last pid: 15892; load averages: 4.51, 4.40, 4.55 11:02:57
173 processes: 153 sleeping, 6 running, 1 zombie, 13 stopped
Cpu states: 49.5% user, 0.0% nice, 50.5% system, 0.0% idle
Memory: 121952K available, 110184K in use, 11768K free, 5872K locked
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
15506 mg 56 0 376K 624K run 5:38 10.95% 10.94% prog.smo9.trac
159 root 55 0 1392K 1008K run 1090:21 10.95% 10.94% automount
28116 root 56 0 280K 464K run 7:43 11.33% 9.77% rpc.netisamd
15891 ssb 57 0 2456K 2128K run 0:02 15.12% 8.20% sas
15599 kml 1 0 456K 200K sleep 853:41 8.21% 8.20% runnewstore
15856 leb 1 0 344K 656K sleep 0:04 6.74% 5.86% returnmiss
15892 jcl 3 0 488K 680K sleep 0:00 7.69% 0.39% se
11539 root 28 0 944K 1152K run 4:04 0.00% 0.00% top
15873 cdt 28 0 416K 408K run 0:00 0.00% 0.00% input
--------------------------------------------------------------------------
--------------------------------------------------------------------------
From: B.Rea@csc.canterbury.ac.nz
Sender: "Bill Rea, University of Canterbury, New Zealand"
<CCTR114@csc.canterbury.ac.nz>
>Is this normal? If not, what can I check to try
>to isolate the problem?
>
> 159 root 55 0 1392K 1008K run 1090:21 10.95% 10.94% automount
Sounds rather screwy to me. To find out what the automount is up to you
can su to root and do a trace -p 159 I would guess that automount has
got itself in a loop. trace should show you what its doing.
---------------------------------------------------------------------------
From: morrow@cns.ucalgary.ca (Bill Morrow)
Was the automounter a major CPU consumer for a long time? It will put
quite a load on temporarily, I'd think, especially if you were doing
a dump across the network as your tape job. That kills NFS response here.
You're right, it shouldn't stay high though.
--------------------------------------------------------------------------
From: srm@shasta.gvg.tek.com (Steve Maraglia)
I have the same problem on numerous systems (SS10 & SS2's). Sometimes
the automount daemon jumps up to %20 percent of the CPU. I've called
Sun about it but they had no solution (That was about 6 months ago).
Please summarize your responses!
--------------------------------------------------------------------------
From: Robert L Krawitz <rlk@Think.COM>
Sender: rlk@Think.COM
Remember that any access through an automounter mount point will cause
the automounter to wake up. This means that a process cd'ed into an
automounted directory will not cause a problem, but repeated stat's that
cross an automountpoint will.
--------------------------------------------------------------------------
From: Christian Lawrence <cal@soac.bellcore.com>
I guess my first question is why are you even running automount on your main
server since it was invented for clients. If you don't need it and/or can live
with a minimum of static mounts, comment it out of the start-up. I say this
because even though automount is a requirement (as far as I'm concerned), it
does have a number of rather nasty bugs associated with it .... especially
prior to 4.1.2. Even then, there are a number of patches not incorporated
as of 4.1.3.
my advice : don't run it on a server unless you absolutely have to **AND**
scan the OSS database for the automount related series
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:28 CDT