Currently running a master and 2 slaves with redis sentinel installed on each redis node.
Running into a rare issue where one of the slaves will drop and then has to re-sync with the master. During this time, sentinel is still sending traffic to the slave with stale data. This is causing massive issues and downtime as we have several services connecting to the slaves to read data. Some go to the healthy/replicating slave and some connect to the stale data slave that is doing a re-sync with the master.
Is there any way to have sentinel mark this slave node as down until its back up and replication is online? Shouldn't sentinel only acknowledge a slave is up if it's state is online on the master and not when in bulk_send status or "down" on the slave?
与恶龙缠斗过久,自身亦成为恶龙；凝视深渊过久,深渊将回以凝视…
