hang w/top, ps commands

colleen hayase

unread,

Mar 5, 2001, 1:27:37 PM3/5/01

to

There was some discussion a while ago regarding system hangs when
running top or a simple "ps -ef" command even though the load is low and
lots of memory is available. Certain processes seem to be hanging the
system. Doing an "ls -l /proc/<PID>" will also hang (if you know the
offending PID). A search of the SunSolve website suggested adding this
line to the /etc/system file:

set kobj_map_space_len = 0x200000

I've used the value "0x800000" and am still running into problems. This
is an Enterprise 6500 machine with 28GB of RAM, 6-400MHz UltraSparc II
processors and 58 GB of swap space. I wondered if the value I'm using is
too low for the machine we have or if there are any other parameters
that should be tweaked. Any info would be greatly appreciated. Thanks!

Colleen

Darren Dunham

unread,

Mar 5, 2001, 7:59:15 PM3/5/01

to

colleen hayase <hay...@us.ibm.com> wrote:
> There was some discussion a while ago regarding system hangs when
> running top or a simple "ps -ef" command even though the load is low and
> lots of memory is available. Certain processes seem to be hanging the
> system. Doing an "ls -l /proc/<PID>" will also hang (if you know the
> offending PID). A search of the SunSolve website suggested adding this
> line to the /etc/system file:

Actually, both ps -ef (but not ps -e) and ls -l (but not ls -ln) make
getpwuid calls to get the username associated with a UID. If something
is blocking those calls (NIS/LDAP server down), then you may see those
symptoms.

If 'ls -l' hangs while 'ls -ln' works, then that's where your problem
lies.

--
Darren Dunham ddu...@taos.com
Unix System Administrator Taos - The SysAdmin Company
Got some Dr Pepper? San Francisco, CA bay area
< Please move on, ...nothing to see here, please disperse >

Roger A. Faulkner

unread,

Mar 6, 2001, 1:25:42 AM3/6/01

to

In article <3AA3DA99...@us.ibm.com>,

colleen hayase <hay...@us.ibm.com> wrote:
>There was some discussion a while ago regarding system hangs when
>running top or a simple "ps -ef" command even though the load is low and
>lots of memory is available. Certain processes seem to be hanging the
>system. Doing an "ls -l /proc/<PID>" will also hang (if you know the
>offending PID). A search of the SunSolve website suggested adding this
>line to the /etc/system file:
>
>set kobj_map_space_len = 0x200000

This has absolutely nothing to do with the problem.
Go and burn the document you found it in.

One common problem that leads to "ps -ef" and "ls -l /proc/<PID>"
hanging is a process that is hung attempting to exec an executable
file from a dead NFS server. The only fix for this is to reboot
the dead server (or fix the ethernet conection to the server, in
case it is not dead but just disconnected somwhow).

Roger Faulkner
roger.f...@sun.com

Pr

unread,

Mar 6, 2001, 5:36:12 AM3/6/01

to

Sometimes, it can be even that the user has some directories in his/her path
which are nfs/auto mounts. If the servers on which the directories are
physically present are down , then some processes on the user's system will
be slow. Check if the nfs mounts are hard/soft mounted?
-Pr

Greg Andrews

unread,

Mar 6, 2001, 9:15:48 AM3/6/01

to

r...@sunraf.Sun.COM (Roger A. Faulkner) writes:
>In article <3AA3DA99...@us.ibm.com>,
>colleen hayase <hay...@us.ibm.com> wrote:
>>There was some discussion a while ago regarding system hangs when
>>running top or a simple "ps -ef" command even though the load is low and
>>lots of memory is available. Certain processes seem to be hanging the
>>system. Doing an "ls -l /proc/<PID>" will also hang (if you know the
>>offending PID). A search of the SunSolve website suggested adding this
>>line to the /etc/system file:
>>
>>set kobj_map_space_len = 0x200000
>
>This has absolutely nothing to do with the problem.
>Go and burn the document you found it in.
>

The document was most likely Infodoc 20267. However, that infodoc's
description makes it clear that it applies to a special set of
commands:

Affected commands include :

prtconf, netstat, dmesg, crash, and adb.

and is not relevant to ordinary commands like ps and ls.

-Greg
--
+++++ Greg Andrews +++ ge...@panix.com +++++
I have a map of the United States that's actual size
-- Steven Wright

Chris Thompson

unread,

Mar 6, 2001, 10:22:02 AM3/6/01

to

In article <981vt6$9ko$1...@engnews2.Eng.Sun.COM>,

Roger A. Faulkner <r...@sunraf.Sun.COM> wrote:
>In article <3AA3DA99...@us.ibm.com>,
>colleen hayase <hay...@us.ibm.com> wrote:

[...]
>> [...]. A search of the SunSolve website suggested adding this

>>line to the /etc/system file:
>>
>>set kobj_map_space_len = 0x200000
>
>This has absolutely nothing to do with the problem.
>Go and burn the document you found it in.

Gee, how do we get to burn down SunSolve? ITWSBT. :-)

Chris Thompson
Email: cet1 [at] cam.ac.uk

Jeffrey Boulier

unread,

Mar 6, 2001, 1:00:05 PM3/6/01

to

In article <DDWo6.100$CY.1...@news.pacbell.net>,

Darren Dunham <ddu...@redwood.taos.com> wrote:
>colleen hayase <hay...@us.ibm.com> wrote:
>> There was some discussion a while ago regarding system hangs when
>> running top or a simple "ps -ef" command even though the load is low and
>> lots of memory is available. Certain processes seem to be hanging the
>> system. Doing an "ls -l /proc/<PID>" will also hang (if you know the
>> offending PID). A search of the SunSolve website suggested adding this
>> line to the /etc/system file:
>
>Actually, both ps -ef (but not ps -e) and ls -l (but not ls -ln) make
>getpwuid calls to get the username associated with a UID. If something
>is blocking those calls (NIS/LDAP server down), then you may see those
>symptoms.
>
>If 'ls -l' hangs while 'ls -ln' works, then that's where your problem
>lies.

This happened to me before, with the username lookups blocked due to a
malfunctioning nscd. Restarting the nscd daemon did the trick on my
system.

Yours Truly,
Jeffrey Boulier
--
Community Source & Support
------=>Prometheus<=------