Re: [Taskforest-discuss] Task Forest Daemon Problem

[Aijaz A. <en...@us...>]

Rosco:

Thanks for the input.  Could you please clarify:  Is it taskforestd (the web
daemon) that goes down or taskforest (the actual engine)?

I'm gonna try duplicating this on my dev box.  When you rerun the job, can
you say for sure that you haven't removed the job that you're trying to
rerun from the Family file?  I know in all versions out there right now,
when you rerun a job that doesn't exist any more, the system doesn't behave
gracefully.  This has been fixed in version 1.34 (been in testing for about
a week).  Another user (Teresia) brought this to my attention in a previous
email to this list.

Is the log directory NFS mounted?  There may be a condition where a file may
not be visible immediately on an NFS mount.

Do you see any errors in your log files (either the taskforest log file or
the taskforestd stdout or stderr log files) when this happens?  If you do,
please forward those errors to me.

I'll try to duplicate this on my dev box this evening and get back to you
after that.

Thanks,

Aijaz.


On Thu, Jun 11, 2009 at 1:45 PM, Rosco Rouse <Ros...@sa...> wrote:

> ...

to rerun a job.  In the olden days, I'd do this with the batch
> "./bin/rerun..." command.  Nowadays, I'm clicking the web client "Rerun"
> button.  All seems well as you issue the command.  Then after several
> minutes the job doesn't appear to finish when I think it should and I begin
> snooping around.  I then discover that the ./bin/taskforest process itself
> has terminated.
>
> Sometimes a single, isolated rerun command will cause this...but not very
> often.  Sometimes, I find it necessary to fire-off several rerun commands
> from a ksh...or lately, with the web client I'll click several jobs as quick
> as the web page can be rebuilt.  The more rerun commands I fire at the
> daemon in a short time, the greater the chance that it'll go down.
>
> Generally, we have very little trouble with the daemon.  It performs
> reliably in all but this circumstance.
> ...
>