LIMIT 154 doesnt seem to work for me, but LIMIT 100 does.
So it seems the behaviour of showing / how showing the labels inconsistent

From Telegram...

In T310234#8227516, @Tarrow wrote:

I think keeping them sufficiently updated is probably the expected fix for this ticket

!213 docker: use docker-compose restart for speed (docker-compose-restart)
https://gitlab.wikimedia.org/repos/releng/cli/-/merge_requests/213

I can only imagine that the match you are getting is https://www.wikidata.org/wiki/Q34433 from Wikidata and that this is one part of the version of cradle that is running on cloud that is not yet wikibase agnostic.

This should also get added to the WMDE offboarding checklists for folks leaving.
CC @conny-kawohl_WMDE

Should this state also contain other details such as what mediawiki instances are installed (e.g. dbtype, dbname, mw install path, port)?

New PR now at https://github.com/wmde/wikibase-release-pipeline/pull/359

What exactly will a complete backup recovery entail?

@DD063520 do you have a minimum example docker-compose file that you might be able to provide for your setup?

https://github.com/wmde/wikibase-release-pipeline/pull/356

Just an image for context and confirming this is even an issue as an admin user.

@Deniz_WMDE I'd also be interested to know the payload of the response to /mine as it looks both cached and broken? as the wiki names are not displayed.

Will have been released in v0.15.0

I just looked at a one off hack solution to try and make this work for @noarave using the follwing

In T310898#8164510, @OlafJanssen wrote:

If $wgWBRepoSettings['statementSections'] is set for cloud wikis, will it be possible to overrule this setting by adding a custom property ordering page such as https://kbtestwikibase.wikibase.cloud/wiki/MediaWiki:Wikibase-SortedProperties ?

+1 better errors could be nice. I believe this is currently based on the same JSON validation for the datavalues as the action API, which has always been a little lacking in its responses to users.

And then with the other patch (caching IpSet) that I wrote applied on top

The loop of 10 is about accurate for how many times this was called for wikibase.cloud it seems.
So I tweaked the benchmarker to actually call the singelton each time the benchmark was called.
See P32507

Yeah, today on my local machine I am not able to get the benchmark to show a speed improvement, however yesterday I did.

So 3-4x the size in cache, in exchange for the speed improvement.
If desired I could probably rework this into a temporary switched mode? We could even try it in wmf production then too?

Crossing linking to T309070: [Timebox: 18hrs] Frequent 502 responses when submitting edits from Wikibase.cloud which is what led me to looking at the caching and performance of IPSet for StopForumSpam.
Also some further investigation in T299522: [INVESTIGATION 16hrs] Speed improvements for saving statements and creating new items specifically T299522#7973128
And the currently still open T313215: [24hrs] Investigate how to set up StopForumSpam correctly

Running with the list used in WMF production, and 50 iterations

Benchmarking using a local list now, I believe previous benchmarks were using the internet download

So that explains the memory difference

For one off cases, and immediate need, the batches that were generated in the platform API during the outage can just be deleted, and the query updater will update all entities that changed during the outage.

This probably needs a reworking of the query updater for wbstack and how it interacts with the platform API to get batches etc.

In T314996#8145344, @Deniz_WMDE wrote:

possible ways I can think of

• deviate from mediawiki/wikibase codebase in form of a fork, which we will have to maintain

It sounds like this would be desirable for all 3rd party wikibases.
If so, this should just be changed in the text strings in wikibase itself.

This certainly doesn't count as the investigation, but just some thoughts.

What's the ticket for finding a solution to the scalability issues?

It would be good to get a list of perceived up and downsides to needing to use the pinned chart repo.

Back when I created wbstack I intentionally disabled custom JS and CSS.
This was mainly because I didn't want to think about the "right" way to configure this at the time balancing security, and usefulness for people, also while trying to not result in people adding 100 gadgets to their wikibases having to maintain them all everywhere etc.

Going to resolve this then, as It looks like everything was happening as expected, just the wrong account was being used.

Sounds like this makes sense!
There is probably a pattern to follow here where the pom file is copied, deps installed, then the rest of the code copied and built?

Ping @Tarrow I'd love to know too :)
But perhaps this is still from before the rewrote the building to use wikiman and pacman?

New text was already done in T306868 for hovering, but not for errors yet it seems.