Page MenuHomePhabricator

Some entities with changed descriptions have old descriptions in WQS and search
Closed, DuplicatePublic5 Estimated Story PointsBUG REPORT

Description

I have changed the Czech descriptions of some Wikidata entities. And later, I have found that a few of these changed entities still show old (pre-edit) descriptions in WQS and in search results on Wikidata.

Take a look at this WQS query which shows the Czech descriptions of four affected entities. All of the displayed descriptions read “rozcestník”. Compare it with one of those entities which in fact has “rozcestník na projektech Wikimedia” (changed in this edit).

The same old description is displayed as a snippet in the full-text search results (API). On the other hand, the search box suggestion dropdown shows the correct updated text (API?)

This is not really a big problem for me. I am not sure if it is worthwhile to create a bugreport for it or it should fix itself in some time (it’s been almost two weeks and it did not fix itself yet). And if it is normal and well-known that WQS sometimes gets out of sync or this is a special terrible bug worth fixing. :-) So – yeah, do anything you consider useful with this report, close it as uninteresting if you want, whatever.

Event Timeline

Gehel subscribed.

Thanks for the report! There are a number of known issues with the current update process for WDQS. We are rolling out a new updater for WDQS that should be out in the next couple of weeks, so we'll not investigate this further on the WDQS side.

The Search indices are updated via a different process, which should be a lot more robust. It is unexpected to not see updates on the Search side, we'll investigate that further.

If this problem happened for both CirrusSearch and WDQS I suspect a problem with changepropagation and/or eventgate.
WDQS have been reloaded and no longer shows the problem.
Search does still have the revision 1412772598 in its index while 1502388913 is expected.

Checking SAL I don't see anything obvious.
Checking logstash for this time period I see nothing.
Checking event.mediawiki_revision_create I can't find this revision either, even scanning one day before and after:

spark.read.table('event.mediawiki_revision_create').where('year = 2021 AND month=9 and day in (23,24,25) and rev_id=1502388913').show(10, False)

shows nothing.

Note that the search index will correct itself at some point, @Mormegil please let me know if you want me to force refresh this specific item.

I'm closing this task as a duplicate of T215001 as I believe this is the root cause.

please let me know if you want me to force refresh this specific item.

That’s not necessary for me, thanks!