-
Have 2 endpoints (master & slave) that receive the same http webhooks Assume hosted at 2 different cloud providers Assume 2 splitted databases, direct joins are not possible They can talk over http, but also expose database connection (though will be 2 separate d
-
Webhooks are stored in a DB for later processing Don’t need to worry about processing, just store Up to you which DB engine you use
-
Master checks every 2 min if it received all webhooks that slave received in interval [-4min,-2min ago] Make an efficient comparison algorithm that doesn’t kill the server or the network (we receive up to 100 webhooks/second → 12k webhooks/2 minutes) Hooks are unique when body is the same + max 1 min diff in received time So “forgotten webhooks” = slave[-4min, -2min] - master[-5min, -1min] E.g.: slave received body “{a:b}” at 19:45:59 UTC, we should not insert into master when it has a body “{a:b}” between 19:44:59 and 19:46:59 UTC.
-
Endpoints that receive json bodies [POST] master.probackup.io/v1/webhooks [POST] slave.probackup.io/v1/webhooks The bodies are always json but do not have a fixed structure Response “RECEIVED” is ok
-
Store webhooks async in batch Do not store events 1 by 1 (synchronously in the web request) but accumulate them and store in batch
-
Database Preferably use AWS DynamoDB