Restbase started hitting ratelimits as a result of clients hitting the transform endpoint:
2021-02-28.log:19:40 <+icinga-wm> PROBLEM - restbase endpoints health on restbase1029 is CRITICAL: /en.wikipedia.org/v1/transform/wikitext/to/html/{title} (Transform wikitext to html) is CRITICAL: Test Transform wikitext to html returned the unexpected status 429 (expecting: 200) https://wikitech.wikimedia.org/wiki/Services/Monitoring/restbase
We can see the spikes in log messages in the 7 day view here https://logstash.wikimedia.org/goto/d40f82dc85d6053c4483fcce357e9236
The volume of log messages caused issues with logstash around the same time:
17:58 <+icinga-wm> PROBLEM - Logstash Elasticsearch indexing errors #o11y on alert1001 is CRITICAL: 65.36 ge 8 https://wikitech.wikimedia.org/wiki/Logstash%23Indexing_errors https://logstash.wikimedia.org/goto/3283cc1372b7df18f26128163125cf45 https://grafana.wikimedia.org/dashboard/db/logstash
When we look at the message from restbase, there is some additional weirdness - x-client-ip is for the IP of the restbase instance itself, which is possibly related to how the serviceproxy behaves: https://logstash.wikimedia.org/app/discover#/doc/logstash-*/logstash-restbase-2021.02.24?id=H0tT1XcBsCn0xdb8Djvf
There's a further weirdness in this issue from the fact that the messages contain differing wikis in the root_req.uri and api_path - I have no idea what that might be related to.