Run all WikiGrok campaigns at 100% for 2 weeks. Depends on https://trello.com/c/XnJdxmUd/3-13-send-wikigrok-response-data-to-its-own-table
Upped the sampling rate to 25% per discussion with Sam and SWAT deployed that today. When I tested it on en.wiki, I realized there's still a lot of important WikiGrok changes that haven't made it to English Wikipedia yet, especially:
https://gerrit.wikimedia.org/r/#/c/201862/
https://gerrit.wikimedia.org/r/#/c/203971/
https://gerrit.wikimedia.org/r/#/c/203312/
In light of that, I decided to postpone turning it on until the next train reaches en.wiki on Wednesday (April 22).
@phuedx, @JKatzWMF This is running now in production. I've verified that the UI is displaying to users and it is collecting data in the wikidatawiki table. All the response data looks correct. Would be good to have Analytics (Leila) verify the EventLogging data as well. I'm leaving for vacation tomorrow, so someone else will need to head that up.
@leila: Is double-checking the EventLogging data for WikiGrok something that I contact you directly about or should I go via the Analytics mailing list?
@phuedx I'm not sure given all the changes that have happened recently. We missed a syncing meeting on Tuesday. I'll call for a 15-min meeting that Jon, Kevin, Dario, Dan, you and figure out the next steps.
As discussed on mobile-l and in-meeting, we'll be rolling back to the last stable WikiGrok deploy this morning/evening (mevening). Consequently, WikiGrok will be using EventLogging to collect data for the new campaigns, which has been signed off already.
@phuedx happy to see this moving forward.
A note on the 25% rate per our discussion on Friday. Please follow the exact same procedure and rate for sampling users as we did in the last stable deploy. Otherwise, EL may not be able to handle the load and the data may not be useful for analysis. If you decide to change the rate, please consult with Analytics. If you decide to change the process you sample, please talk to Dario or myself.