[daip] AIPS 1: TASK ACTIVE problem

[Philip Lah]

Hello Eric,

   the experiments I did with DOWAIT true were back in 2004 and I cannot 
remember now exactly why I decided against using it.  My code currently 
logs into using AIPS remote with the message text piped to a text file. 
The perl script checks this file every second for 'Appears to have ended 
successfully' and only then does it continue to the next step.  It also 
checks this file for a variety of other text indications that something 
goes wrong like 'TASK ACTIVE' and if it finds it the script exits.  The 
reason I chose not to use DOWAIT was was either because the perl script 
didn't wait for AIPS to finish despite the DOWAIT or because the perl 
script did not pick up when AIPS had a problem with the AIPS task just 
run.  I will look in to having the script checking for the FITTP1.$nnnnn 
file so that my data reduction run does not randomly stop half way through 
the 2+ days of processing.

   Bye
        Philip Lah

On Wed, 2 Nov 2011, Eric Greisen wrote:

> Philip Lah wrote:
>> Hello Eric,
>> 
>>   unfortunately I do not have much I can add to give a more complete 
>> description of the problem.  I am using perl scripts to interface with AIPS 
>> for my data reduction.  For some reason on the new machine I now 
>> occasionally get the TASK ACTIVE halting my task and stopping my script. 
>> Years ago I did look at using DOWAIT but I discarded it as an option 
>> because it did not give the result I wanted with my scripts.
>> 
>>   From your email I now have a suspiscion about what may be going wrong. In 
>> the perl script that halted prematuraly last night I have multiple FITTP 
>> commands that run one after another outputting the 25 facets of my 
>> continuum image.  I am guessing the for one of the runs the FITTP1.$nnnnn 
>> for the precious run in the /tmp dir had not been removed before the next 
>> FTTP started, so AIPS decided to halt with the TASK ACTIVE error.  This 
>> does not always happen as I ran the same script today with no trouble.  I 
>> suspect that the machine was under higher read/write pressure when the 
>> problem occured hence the delay in removing the old FITTP1.$nnnnn.  I am 
>> not sure how to stop this problem from happening again.  Perhaps I should 
>> have my scripts check for the existance of the FITTP1.$nnnnn before 
>> deciding to continue.
>
> Actually you should probably set DOWAIT true or before running a task give 
> the command
>     WAIT <task>
> where <task> is the task name.  The WAIT will finish immediately if the task 
> is not running and will wait for it to finish otherwise.  You are
> running a script depending on race conditions and that is unwise.  The
> DOWAIT true does allow the task to decide that it has failed and then notify 
> AIPS while otherwise all the task knows is that it started up okay.
>
> Eric Greisen
>

**************************************************
     Philip Lah
   Postdoctoral Fellow, joint appointment at
   National Centre for Radio Astrophysics
     and
   Australian National University

   Home Page: http://www.mso.anu.edu.au/~plah/

**************************************************