[GE users] 6.2u3 qsub Segmentation fault when using -V and -jsv ?

victorg vgregorio at penguincomputing.com
Tue Sep 15 16:43:34 BST 2009


Hello Mark,

I tried, but cannot reproduce your problem on CentOS 5.3 running 6.2u3 
lx24-amd64.  If you can produce a core dump, what does the backtrace in 
gdb look like?

[vgregorio at scyld ~]$ which qsub
/opt/sge/bin/lx24-amd64/qsub

[vgregorio at scyld ~]$ qsub -V test.sh
Your job 7 ("test.sh") has been submitted

[vgregorio at scyld ~]$ qsub -jsv ./test_pl.jsv test.sh
Your job 8 ("test.sh") has been submitted

[vgregorio at scyld ~]$ qsub -V -jsv ./test_pl.jsv test.sh
Your job 9 ("test.sh") has been submitted

Regards,

--
Victor Gregorio
Penguin Computing


On 09/15/2009 06:42 AM, ccaamad wrote:
> Hi,
>
> I'd appreciate it if someone could see if they can reproduce this before I
> submit it as a bug report.
>
> Continuing my exploration of 6.2u3, I've been looking at the JSV
> functionality. I love it - I can do all sorts of things without asking
> users to change how they submit their jobs (like selecting PE depending
> on topology and number of processes).
>
> However, on my CentOS 5.3 lx24-x86 VMware test box, I see that with even a
> simple client JSV, "qsub -V" suffers a segmentation fault :(
>
> Can anyone reproduce this on lx24-x86 or lx24-amd64 architectures, please?
> I really hope I've just done something wrong!
>
> e.g.
>
>     $ qsub -V test.sh
>     Your job 183 ("test.sh") has been submitted
>
>     $ qsub -jsv ./test_pl.jsv test.sh
>     Your job 184 ("test.sh") has been submitted
>
>     $ qsub -V -jsv ./test_pl.jsv test.sh
>     Segmentation fault
>
> Doing this as a system JSV has an even worse effect - it seems to actually
> kill the batch system processes:
>
>     $ qconf -sconf | grep jsv_url
>     jsv_url                      /services/ge6.2u3/jsv/test_pl.jsv
>     $ qsub test.sh
>     Your job 187 ("test.sh") has been submitted
>     $ qsub -V test.sh
>     error: commlib error: got read error (closing
>     "s1.issmcd.leeds.ac.uk/qmaster/1")error: commlib error: got select error (Connection refused)
>     Unable to run job: unable to send message to qmaster using port 802 on host "s1.issmcd.leeds.ac.uk": got send error.
>     Exiting.
>
> Contents of files:
>
> $ cat test_pl.jsv
> #!/usr/bin/perl
>
> use strict;
>
> use lib "$ENV{SGE_ROOT}/util/resources/jsv";
> use JSV qw(:ALL);
>
> jsv_on_start(sub {
>      jsv_send_env();
> });
>
> jsv_on_verify(sub {
>       jsv_accept();
>       return;
> });
>
> jsv_main();
>
> $ cat test.sh
> echo hello
>
> This is a bit of a problem: jsv_set_param("V","y") doesn't seem to get
> round this by doing it in the JSV.
>
> Thanks,
>
> Mark

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=217347

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list