Page MenuHomePhabricator

Spike in ResourceLoader warnings after rolling wmf.10 to group1 wikis
Closed, InvalidPublicPRODUCTION ERROR

Description

There was a noticeable spike in warnings from resource loader after rolling out 1.27.0-wmf.10 to group1 wikis:
https://logstash.wikimedia.org/#dashboard/temp/AVI8iTvlptxhN1XabrfX

Event Timeline

thcipriani raised the priority of this task from to Needs Triage.
thcipriani updated the task description. (Show Details)
thcipriani subscribed.
Krinkle claimed this task.
Krinkle subscribed.

The spike is just a temporary amplification of the normal "Message x is not defined" warnings which have existing tasks filed already.

They are amplified after deployment because deployment rebuilds localisation cache which triggers a revalidation of ResourceLoader message blobs. This is intentional and ensures that updates messages are actually deployed. As part of this, it will also walk over messages that turn out not to exist. These warnings are mostly dormant at other times because messages are strongly cached.

Closing this task because it essentially just says "Errors happen when broken code is triggered." – where the relevant code here is code triggered by deployment. (Not this deployment in particular, but any.)

Example of a task for the broken code itself:
T118820: MessageBlobStore::fetchMessage failed to find: 'multimediaviewer-option-header-viewer' (and other multimediaviewer keys)

mmodell changed the subtype of this task from "Task" to "Production Error".Aug 28 2019, 11:11 PM