Re: Jobs are killed immediately when "ulimit -s unlimited" is in effect (AIX 5.2)

From: Peter S. Shenkin (shenkin_at_mindspring.com)
Date: 06/18/04


Date: 18 Jun 2004 04:28:40 -0700

mky@talk21.com (mark taylor) wrote in message news:<dee0a7c3.0406160355.5b471088@posting.google.com>...
> FYI, IBM does not support bash, the bash shell shipped with 5L AIX
> tool box for linux apps is supplied AS-IS, so I wouldnt go asking them
> to.
>
> Better off building a testcase in ksh or csh if you also see the
> problem there.

Good thinking; though I did say that it occurs in csh. It also
occurs in a program that uses getrlimit/setrlimit, but that
program was likely invoked from bash. Results in various
environments given below.

> Also, why give up, if you have paid for support from
> IBM and you box is broke, then press them to fix it, rasie a critsit
> if you have to etc... its not rocket science... get what you paid
> for... is it actually IBM support or a business partner ? or some 3rd
> party support contract ?

These are loaner machines which we obtained from IBM in order
to support our software under AIX. The advice to reinstall came
from our IBM-designated technical contact.

> do you see the problem in ksh ?

Good thinking. We see a diagnostic -- not exactly the
same problem:

: ./try.ksh[2]: ulimit: outside the allowable range
: 2097152
: hello, world 1
: 2097151
: hello, world 2

The diagnostic is invoked by "ulimit -s unlimited". I added
the "1" and "2" to the printout for clarity.

> or just csh and bash ?
>
> whats your csh test case ?

Similar to the bash/sh/ksh test cases, except using "limit stacksize"
wherever I previously used "ulimit -s". This gives:

: stacksize 4194304 kbytes
: Killed
: stacksize 2097151 kbytes
: hello, world 2

Note that the result of setting stacksize to "unlimited" is different
here.

For completeness, I'll say that sh gives the same result as ksh.

> what level of code are you on ? bos.mp?
: bos.mp 5.2.0.10 5765-E6200 bos.mp

> bos.rte.shell ?
: bos.rte.shell 5.2.0.10
: 5765-E6200 0000 bos

Further thoughts appreciated. I'll certainly go back to my
IBM contact with the above, in the meantime....

-P.



Relevant Pages

  • Re: IBM extends support for AIX 5.1.
    ... I don't totally agree (not with you, but rather your customers ... IBM got the word out long ago about their direction on ... AIX Administrator, Komatsu America ... is because they are withdrawing support of 4.3.3. ...
    (AIX-L)
  • Re: IBM extends support for AIX 5.1.
    ... any RS/6000 machine and that was a selling feature for IBM. ... the support deadline. ... AIX Administrator, Komatsu America ... that there are some fairly recent machines that will not run 5.2. ...
    (AIX-L)
  • Re: IBM extends support for AIX 5.1.
    ... AIX Administrator, Komatsu America ... any RS/6000 machine and that was a selling feature for IBM. ... the support deadline. ... that there are some fairly recent machines that will not run 5.2. ...
    (AIX-L)
  • Re: Default Catch on 43p-240
    ... > prep machine. ... The reason why IBM called it RSPC was because they were trying to ... for AIX 4.3.3 on December 31 of this year. ... but its support ends in 2004 or so. ...
    (comp.unix.aix)
  • Re: Jobs are killed immediately when "ulimit -s unlimited" is in effect (AIX 5.2)
    ... FYI, IBM does not support bash, the bash shell shipped with 5L AIX ... Better off building a testcase in ksh or csh if you also see the ...
    (comp.unix.aix)