Page MenuHomePhabricator

cswiki beta: Remove Add Image tasks that do not have suggestion data in production
Open, LowPublic

Description

During the investigation of T377783: [cswiki betalabs] Structured tasks do not load editing surface , I noticed a significant proportion of Add Image tasks in cswiki beta are invalid, as cswiki production no longer has suggestion data for the corresponding production article. This issue happens, because Beta does not have its own source of Add Image tasks. Instead, it reuses production data, because the actual Add Image service cannot be used from outside of production (and even if it could be, it would still provide data about production wikis, not beta articles).

To make testing and QA easier, we should refresh the set of Add Image tasks at cswiki beta. This needs to be done via the extensions/CirrusSearch/maintenance/UpdateWeightedTags.php maintenance script, manually specifying which articles should be removed from the index and which articles should be added. We can either remove all articles and generate a fresh set of articles from production, or somehow attempt to remove only the broken suggestions.

We should also consider doing this change for any other beta wiki where we regularly need Add Image to work, but given this is quite a lot of manual work, we should strive to keep the set of wikis to the bare minimum. For now, filling this task as cswiki beta specific.

Event Timeline

@Etonkovidova @KStoller-WMF: Do you think we need working Add Image on other (beta) wikis besides cswiki beta? If so, which ones? Maybe pilots?

@KStoller-WMF This issue will keep happening, as production data is changing, and recommendations are aging as well. We should probably invest into improving the behaviour rather than doing some manual work every now and then – do you have thoughts regarding prioritization here?

Do you think we need working Add Image on other (beta) wikis besides cswiki beta? If so, which ones? Maybe pilots?

Yes, perhaps we should do this for eswiki beta and frwiki beta since we plan to pilot test Surfacing Structured Tasks work (T362584) on those two wikis. Should we just update this task to include them as well?

This issue will keep happening, as production data is changing, and recommendations are aging as well. We should probably invest into improving the behaviour rather than doing some manual work every now and then – do you have thoughts regarding prioritization here?

Hmmm, I think this depends on how much effort it would be to somehow automate repopulating beta wiki suggestions. If we think we could do that in a way that isn't overly complex or prone to breaking, then I'm all for trying to prioritize it in Q2 or Q3. But if it's a significant effort, we might want to continue to just manually do this when we noticed it's needed.

Do you think we need working Add Image on other (beta) wikis besides cswiki beta? If so, which ones? Maybe pilots?

Yes, perhaps we should do this for eswiki beta and frwiki beta since we plan to pilot test Surfacing Structured Tasks work (T362584) on those two wikis. Should we just update this task to include them as well?

+1 to have ADd image on pilot wikis.

This issue will keep happening, as production data is changing, and recommendations are aging as well. We should probably invest into improving the behaviour rather than doing some manual work every now and then – do you have thoughts regarding prioritization here?

Hmmm, I think this depends on how much effort it would be to somehow automate repopulating beta wiki suggestions. If we think we could do that in a way that isn't overly complex or prone to breaking, then I'm all for trying to prioritize it in Q2 or Q3. But if it's a significant effort, we might want to continue to just manually do this when we noticed it's needed.

I was wondering if I could take on running the script on beta and do any other manual work that needed.