Indexing cycles repeated; batch job restarts automatically

• Article Type: General
• Product: Aleph
• Product Version: 15.2

Description:
p_manage_02 for abc01 has been running for several days. (It should take only one day.) We find that it is *repeating* cycles: the $data_scratch/p_manage_02.cycles file shows that cycle 6 is done, but when we look at it the next day, it cycle 6 shows as *not* done.

Resolution:
After p_manage_02 for abc01 finishes, it is automatically restarting.

This can be seen in the abc01 $data_files/lib_batch_log file:

abc01_p_manage_02.00011
Thu Jan 04 18:35:58 aleph *** ************************************************
Thu Jan 04 18:35:58 aleph *** executing /exlibris/aleph/a17_7/aleph/proc/p_manag
e_02 with parameters ...
Thu Jan 04 18:35:58 aleph *** log file is: /exlibris/aleph/u17_7/alephe/scratch/
abc01_p_manage_02.00011
Fri Jan 05 19:30:18 aleph *** ************************************************
Fri Jan 05 19:30:18 aleph *** executing /exlibris/aleph/a17_7/aleph/proc/p_manag
e_02 with parameters ...
Fri Jan 05 19:30:19 aleph *** log file is: /exlibris/aleph/u17_7/alephe/scratch/
<etc.>

Note that despite these multiple runs, we see only one log ("abc01_p_manage_02.00011") in $alephe_scratch. (It keeps overwriting itself when it restarts.)

Why is it restarting? I see the following line in abc01 util c/7 ("Show Queue of Jobs Awaiting Execution"):

p_manage_02 00011 Holding until 20070103 18

I tried deleing this with util c/8, but got the response "Can't delete entry 11". Therefore, I did the following to re-initialize the que_batch file:

dlib abc01
df1
rm que_batch_lock
rm que_batch
touch que_batch

Version 18 rep_change 560 is relevant:

Description: If the batch queue is started when there are jobs' waiting for execution in the queue, the last job line in the queue is duplicated and the job is submitted in an infinite loop. This scenario appears on AIX machines only.

AIX sites not yet on version 18 may want to follow a policy of stopping the batch queue only when it is empty. If you need to stop it when there are jobs in it, then you will need to delete/recreate it as described above, before restarting it.