OPAC performance problems after large MARC loads
- Article Type: General
- Product: Aleph
- Product Version: 20
Problem Symptoms:
After large MARC record loads
1. system waits ("wa" in "top") increase to 35%
2. OPAC searches become extremely slow.
Direct links to records (such as are found on our video lists -- http://libweb.lib.abc.edu/exlibris/aleph/a21_2/tmp/list-dvd-film.htm) work very quickly. It's just searches in the OPAC that fail.
If the xxx01 ue_01 is stopped, the "waiting for IO" percentage goes down to 3-4% and everything works normally.
Cause:
Uncertain.
Resolution:
Wait until later in the day when activity is lower to restart ue_01. It should process the records by morning without any problem.
Additional Information
There were 7603 z07a records to be processed (by ue_01_z0102 and ue_01_word):
xxx01@ALEPH21> select count(*) from z07a;
**** Hit return to continue ****
COUNT(*)
----------
7603
io-waits were around 30%. After stopping/killing the ue_01 processes, the waits decrease to 5-10%.
The ue_01 is processing the records quite quickly.
It could be that these particular bib records are exhibiting this problem while others would not.
If not having ue_01 running during the day and the indexes for the records catalogers are entering in during the day not being updated is a concern, please consult KB 8192-2329 ("Indexes updated immediately; indexes updated by ue_01"). As noted there, the z11 Direct indexes are updated independently of ue_01, so catalogers *should* be able to find the records entered today by OCLC# and ISBN -- just not title.
Category: Background processing (500)
Subject: Ue (500)
- Article last edited: 10/8/2013