• Article Type: General
• Product: Aleph
• Product Version: 18.01

Description:
After upgrading to version 18, bib records are not being indexed.

The vcc01 util c/1 showed the following:

aleph 91894 1 0 Nov 18 - 0:02 /exlibris/aleph/a18_1/aleph/exe/lib_batch VCC01
aleph 93950 1 0 Nov 18 - 0:16 /exlibris/aleph/a18_1/aleph/exe/rts32 ue_01_z0102_index VCC01.a18_1
aleph 94704 1 0 Nov 18 - 0:14 /exlibris/aleph/a18_1/aleph/exe/rts32 ue_01_word_index VCC01.a18_1
aleph 100930 1 0 Nov 18 - 0:03 /exlibris/aleph/a18_1/aleph/exe/rts32 ue_08_a VCC01.a18_1 C

You will note that the main ue_01 process (ue_01_a) was missing.

Looking at the $data_scratch/run_e_01.87808 log, I saw the following:

HANDLING DOC NO. - VCC01.001477922 03:32:07
...
Update z98 from z980 : 001000112022
...
Load error : file 'uncompress_data_ads'
error code: 105, pc=0, call=111, seg=0
105 Memory allocation error

I stopped/restarted ue_01 and then saw the following (in the run_e_01.62732 log):

HANDLING DOC NO. - VCC01.001477923 11:36:54
...
Update z98 from z980 : 001000207292
Update z98 from z980 : 002000228917
Update z98 from z980 : 001000268664
...
Load error : file 'update_long_heading'
error code: 105, pc=0, call=136, seg=0
105 Memory allocation error

v18 rep_ver 11619 shows the following:

Description: Error 105 "Memory allocation error" appears in ue_01 log. This was fixed.
Error message 105 "Memory allocation error" appeared in the ue_01 log file on AIX platforms. This has been corrected.

./alephm/source/com/update_long_heading.cbl

Since we are running AIX, this could be a variation on this previous problem, though it seems that other AIX sites have moved to v18 without any problem.

Here is a performance troubleshooting tips for run-time memory:
If your system is running low on swap space or you have exceeded the maximum data size allowed for a process, you may see the following error messages. In each case you may have to configure your system with more swap space or increase the value of kernel parameter maxdsiz.
105 Memory allocation error
...

Resolution:
From site:

The data stack size was whatever default IBM uses for a new account, which the R 18 user (ALEPH18) was, i.e., a new account. Poking around on the Internet, I see a number of applications, e.g., ORACLE, suggest using -1 which is unlimited in the AIX world.

I changed ALEPH18 to unlimited and the index is running correctly now. I am going to check with the IT folks who know more about AIX than I do and see what the implications of unlimited are.

It looks like the improvements in the indexing uses a lot more memory. At one point when I was messing around, I had topas runing and when I started up the indexing, the system page fault values jumped to over 20,000 faults immidiately, it was trying to get a bunch-o-memory.

• Article last edited: 10/8/2013