Page MenuHomePhabricator

Abstract Wikipedia is not compatible with new API rate limits
Closed, ResolvedPublicBUG REPORT

Description

Steps to replicate the issue (include links if applicable):

  1. Log out (or open a new private tab)
  2. Visit https://abstract.wikipedia.org/wiki/Q188815
  3. Let the page finish loading
  4. Repeat steps 2 and 3 about 7-10 times.

What happens?:

"Unable to render this fragment due to an unknown error." and 429 error in console

What should have happened instead?:

No 429 error

It seems AbWiki pages make one API request per statement. That page has over 100 statements. And it's really not all that big. What happens when a big page like https://en.wikipedia.org/wiki/United_States is imported?

And of course, this doesn't just break AbWiki; the user is now locked out of all API features on all wikis. Including, for example, DiscussionTools, when they go to ask for help.

Event Timeline

A_smart_kitten subscribed.

(as I believe that MWP may be the responsible team to tag for tasks about these new rate-limits, and as the rule to auto-tag Abstract Wikipedia team on these sorts of tasks seems like it may not currently be working)

Change #1265333 had a related patch set uploaded (by Daniel Kinzler; author: Daniel Kinzler):

[operations/deployment-charts@master] rest gateway: introduce policy for abstractwiki/wikifunctions

https://gerrit.wikimedia.org/r/1265333

Change #1265568 had a related patch set uploaded (by Genoveva Galarza; author: Genoveva Galarza):

[mediawiki/extensions/WikiLambda@master] Extend queue processing times for abstract fragments

https://gerrit.wikimedia.org/r/1265568

Change #1265568 merged by jenkins-bot:

[mediawiki/extensions/WikiLambda@master] Extend queue processing times for abstract fragments

https://gerrit.wikimedia.org/r/1265568

Change #1266219 had a related patch set uploaded (by Jforrester; author: Genoveva Galarza):

[mediawiki/extensions/WikiLambda@wmf/1.46.0-wmf.22] Extend queue processing times for abstract fragments

https://gerrit.wikimedia.org/r/1266219

Change #1266219 merged by jenkins-bot:

[mediawiki/extensions/WikiLambda@wmf/1.46.0-wmf.22] Extend queue processing times for abstract fragments

https://gerrit.wikimedia.org/r/1266219

Mentioned in SAL (#wikimedia-operations) [2026-04-01T14:08:28Z] <jforrester@deploy1003> Started scap sync-world: Backport for [[gerrit:1266190|MemcachedWrapper: Hash key when longer than 250 characters]], [[gerrit:1266219|Extend queue processing times for abstract fragments (T421581)]]

Mentioned in SAL (#wikimedia-operations) [2026-04-01T14:10:26Z] <jforrester@deploy1003> jforrester: Backport for [[gerrit:1266190|MemcachedWrapper: Hash key when longer than 250 characters]], [[gerrit:1266219|Extend queue processing times for abstract fragments (T421581)]] synced to the testservers (see https://wikitech.wikimedia.org/wiki/Mwdebug). Changes can now be verified there.

Mentioned in SAL (#wikimedia-operations) [2026-04-01T14:16:43Z] <jforrester@deploy1003> Finished scap sync-world: Backport for [[gerrit:1266190|MemcachedWrapper: Hash key when longer than 250 characters]], [[gerrit:1266219|Extend queue processing times for abstract fragments (T421581)]] (duration: 08m 14s)

Change #1265333 merged by jenkins-bot:

[operations/deployment-charts@master] rest gateway: introduce policy for abstractwiki/wikifunctions

https://gerrit.wikimedia.org/r/1265333

Change #1267122 had a related patch set uploaded (by Daniel Kinzler; author: Daniel Kinzler):

[operations/deployment-charts@master] rest gateway: introduce policy for abstractwiki/wikifunctions"

https://gerrit.wikimedia.org/r/1267122

ftr, the patch didn't go out, we have to sort out the gateway routing for wikifunctions first.

Change #1269079 had a related patch set uploaded (by Jforrester; author: Jforrester):

[mediawiki/extensions/WikiLambda@master] AbstractPreview: Use exponential backoff when retrying renders

https://gerrit.wikimedia.org/r/1269079

Change #1267122 merged by jenkins-bot:

[operations/deployment-charts@master] rest gateway: introduce policy for Abstract Wikipedia

https://gerrit.wikimedia.org/r/1267122

Change #1269079 merged by jenkins-bot:

[mediawiki/extensions/WikiLambda@master] AbstractPreview: Use exponential backoff when retrying renders

https://gerrit.wikimedia.org/r/1269079

daniel claimed this task.

should be fixed now

Escargot_rouge subscribed.

I cannot get https://abstract.wikipedia.org/wiki/Q90 to properly load. I always see "Reached max retries. Try again later."

image_2026-04-16_112124534.png (459×980 px, 41 KB)

It used to work fine for me before rate limiting was enforced.

Redmin subscribed.

I cannot get https://abstract.wikipedia.org/wiki/Q90 to properly load. I always see "Reached max retries. Try again later."

image_2026-04-16_112124534.png (459×980 px, 41 KB)

It used to work fine for me before rate limiting was enforced.

It works after clicking 'Retry'; see T420630.

I cannot get https://abstract.wikipedia.org/wiki/Q90 to properly load. I always see "Reached max retries. Try again later."

@Jdforrester-WMF I suppose that's because you reduced the number of retries?

It used to work fine for me before rate limiting was enforced.

I think that's a coincidence - the retries here are not because of rate limits, but relate to wikifunctions evaluation in the background.

I cannot get https://abstract.wikipedia.org/wiki/Q90 to properly load. I always see "Reached max retries. Try again later."

@Jdforrester-WMF I suppose that's because you reduced the number of retries?

Not quite, the initial failure is because of T422299: Wikifunctions client wikis' cache config doesn't know about wf-wan, so embedded Functions’ output is never available which means fragments aren't getting cached (so every page view is a render), and the service can't run them fast enough.

The reduced, slower run rate means we are now much less likely to hit the rate limits (alongside your changes to isolate our requests), and once the content storage is fixed should restore functionality.

Sorry for the disruption!