Deploy the related pages feature to mobile web stable for English language to 90% of users
NOTE: ensure sampling rate of schema:relatedarticles is within acceptable bounds for eventlogging
NOTE: schema should run for both control and on group
Description
Description
| Status | Assigned | Task | ||
|---|---|---|---|---|
| Resolved | ovasileva | T154681 Deploy related pages to mobile web stable for 90% English | ||
| Resolved | Jdlrobson | T155079 Deploy related pages to mobile web stable to 100% of users for English | ||
| Resolved | Jdlrobson | T157165 Deploying related pages broke our browser tests | ||
| Resolved | ovasileva | T157372 [SUBTASK] RelatedArticlesEnabledSamplingRate breaks RelatedArticles browser tests |
Comment Actions
We cannot do 90% rollouts on related articles. We'd need to add some scaffolding to support this...
Comment Actions
wmf10 will go out to English Wikipedia today meaning we can make use of T156039: Allow RelatedPages to be enabled for subset of users.
Comment Actions
Change 335686 had a related patch set uploaded (by Jdlrobson):
Related pages is shown to 90% of mobile users
Comment Actions
Mentioned in SAL (#wikimedia-operations) [2017-02-03T00:45:07Z] <dereckson@tin> Synchronized wmf-config/: Adjust RelatedArticles deployment scale for Mobile English Wikipedia (T154681) (duration: 00m 42s)
Comment Actions
Mentioned in SAL (#wikimedia-operations) [2017-02-03T00:51:17Z] <dereckson@tin> Synchronized dblists/related-articles-footer-blacklisted-skins.dblist: Adjust RelatedArticles deployment scale for Mobile English Wikipedia (T154681) (duration: 00m 39s)
Comment Actions
Related pages will show to 90% of English Wikipedia users.
Bucket yourself $.cookie('enwikimwuser-sessionId', 'a1') to not see it.
I haven't been able to verify the EventLogging is working correctly or whether we can test the data we collect. Looking at https://meta.wikimedia.org/wiki/Schema:RelatedArticles we only log the session id.. was that an oversight?
Note we can work out the bucket via a script that converts session id to bucket if necessary. Just open a new task for that analysis and give one of the developers a dump of the data you want to convert in csv format.
( we can run mw.experiments.getBucket( x, sessionId ); on each result )
Comment Actions
@Jdlrobson - ok, now onto eventlogging - we're only logging the session id, meaning we're not logging the bucket?
This comment was removed by ovasileva.
This comment was removed by ovasileva.
Comment Actions
@Jdlrobson - almost done! I wanted to double-check if we are currently running the schema on all users
Comment Actions
We talked about this in standup today. We should be running the schema on all users but someone will need to check the database to see if this is actually working. That said, we should be clear what the hypothesis is that we are trying to measure and check that we can actually measure this with the limited schema we have.
Comment Actions
Will track instrumentation in T157375: Update related pages schema to allow for current A/B test. That said, closing this!