1

I have an app with hundreds of horizontally scaled servers which uses redis pub/sub, and it works just fine.

The redis server is a central point of failure. Whenever redis fails (well, it happens sometimes), our application falls into inconsistent state and have to follow recovery process which takes time. During this time the entire app is hardly useful.

Is there any messaging system/framework option, similar to redis pub/sub, but with redundancy and high availability so that if one instance fails, other will continue to deliver the messages exchanged between application hosts?

Or, better, is there any distributed messaging system in which app instances exchange the messages in a peer-to-peer manner, so that there is no single point of failure?

roman
  • 892
  • 9
  • 26
  • What about the Redis cluster setup with sentinel? – sonus21 Feb 23 '21 at 03:54
  • Looks interesting, is the node switch in sentinel fully transparent or will there be some downtime if a master goes offline? In any case this is something we will try soon. – roman Feb 23 '21 at 20:31
  • The node switch is transparent but your application should have the logic to disconnect and connect to the new node. – sonus21 Feb 24 '21 at 03:17

0 Answers0