Closed rowanseymour closed 1 month ago
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 100.00%. Comparing base (
17b8923
) to head (fa2c9be
).
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
We've been doing broadcasts to nodes wrong for a while now, fetching the recipient contact ids, then saving them on the broadcast, then queueing that broadcast. It recently became noticeable because I moved the broadcast creation part into mailroom, and now we request timeout errors when someone tries to such a broadcast to a lot of contacts, because the request from RP to mailroom times out. This approach also puts potentially millions of contact ids into the task payload in Redis. Makes more sense to save the node UUID on the broadcast as part of its recipients, and then let the broadcast send task resolve that when it's resolving contact ids from queries, groups etc.