Hi Jesse,
Yes, the recalculation at the end of that upgrade script is necessary.
In 3.0, we made some changes to the way catalog searches determine
record visibility, and this part of the script recalculates visibility
to fix a few search issues that were discovered in the 3.0 release.
Without recalculating visibility, you'll find that some records for
electronic resources or those that have a bib source (which cover almost
all records in our system) will appear in searches when they shouldn't.
Having said that, I think we can speed up that upgrade script. We had a
similar calculation in the 2.12 to 3.0 upgrade script, and I one point
we made a change to disable various triggers before performing the
calculation. My understanding is that the calculations perform much more
quickly with those triggers disabled. See the changes at:
http://git.evergreen-ils.org/?p=Evergreen.git;a=blobdiff;f=Open-ILS/src/sql/Pg/version-upgrade/2.12.5-3.0-beta1-upgrade-db.sql;h=7fc9b51936854db32a1a09a20ea276bb1a16747e;hp=97ca7fa5fff4bd301dc021cf5a0bac0112a2463b;hb=d388f7019a90a5809514407d7139eb1ed1843432;hpb=0b749e554c3a5c8a93ca36e06e8b587991ab70a3
I'm going to file a bug to see if we can make a similar change for the
3.0.2-3.0.3 upgrade script.
Kathy
On 03/23/2018 03:16 PM, Jesse McCarty wrote:
Hello Everyone,
During my last test cycle we ran into an issue upgrading from 2.10 to
a newer version with an update script that was setting the 901$sfor
bib records. This took an extended amount of time to complete. Well
now, in testing our upgrade to the 3.0 series part of the 3.0.2-3.0.3
version upgrade script took over a week to finish in testing, which is
a big issue for updating production.
Is it possible to comment out/remove the offending part of the upgrade
script and not have any issues with the new system after the upgrade?
Could it be the last part of the script in lines 277-291 of the
upgrade script taking this long (line 290 perhaps)?
277 UPDATE biblio.record_entry
278 SET vis_attr_vector =
biblio.calculate_bib_visibility_attribute_set(id)
279 WHERE id IN (
280 SELECT DISTINCT cn.record
281 FROM asset.call_number cn
282 WHERE NOT cn.deleted
283 AND cn.label = '##URI##'
284 AND EXISTS (
285 SELECT 1
286 FROM asset.uri_call_number_map m
287 WHERE m.call_number = cn.id
288 )
289 UNION
290 SELECT id FROM biblio.record_entry WHERE source IS NOT
NULL
291 );
Wondering if others have met something similar and how they dealt with
it so as not to cause issues upgrading a production system and
minimizing down time. We typically run our upgrades on a Sunday
morning and all Evergreen related services are only down for about
half a day and usually back up before 10am Monday worst case.
Thanks in advance,
Jesse McCarty
City of Burlington
Information Systems Technician
--
Kathy Lussier
Project Coordinator
Massachusetts Library Network Cooperative
(508) 343-0128
[email protected]
Twitter: http://www.twitter.com/kmlussier
