Page MenuHomePhabricator

[Epic] Create a cross-wiki watchlist prototype on the Beta Cluster
Open, Needs TriagePublic

Description

Per T126641 and T142538, we need to set up a cross-wiki watchlist prototype on the Beta Cluster. This will require setting up aggregated watchlist and recentchanges tables that store the relevant data from all wikis.

Event Timeline

kaldari created this task.Oct 23 2016, 10:16 PM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptOct 23 2016, 10:16 PM

@jcrespo: A few questions about setting up the new tables on Beta Cluster:

  • Would it make sense to set up the new tables in the wikishared database or should we create a new database for this?
  • If a new database makes more sense, which shard should that database live on?
  • Are the db shards on Beta Cluster set up the same as in production?
jcrespo added a subscriber: dduvall.EditedOct 24 2016, 8:34 AM

@jcrespo: A few questions about setting up the new tables on Beta Cluster:

I am not the owner at Beta, some people at RelEng is, so make sure to add the relevant people here. Maybe @dduvall or someone else can answer you better. The following are to best of my knowledge, but are not authoritative in any case:

  • Would it make sense to set up the new tables in the wikishared database or should we create a new database for this?

Don't add things to wikishared. This will be on production on a separate set of servers, so it cannot be on the same database as the others, because it will be physically separated.

  • If a new database makes more sense, which shard should that database live on?

If this was production, I would call this x2.

Are the db shards on Beta Cluster set up the same as in production?

I doubt it, I think there is only 2 db servers on Beta. Even if x1 existed, it would probably point to the same servers, which means replication will be difficult to setup on Beta. Replication from localhost seems like a really bad idea. If this want to really test the functionality, I would suggest to create, even if temporarily, at least a new VM for it. Once we know it works, we can simplify the setup, but I would strongly suggest starting with a new hosts only for it (not because it is dangerous/requires resources, but because it will heavily depend on replication to work).