In this experiment, we simulated a failure on the Redis master server 192.168.0.33 by shutting it down manually. The Sentinels promoted 192.168.0.31 as the new master and finished the failover.
Let's see how the entire process happened:
- On Sentinel-1, at 17:05:02.446, it found the master server was unreachable. As we mentioned in the previous recipe, each Sentinel will ping the Redis master, slaves, and other Sentinels regularly. If a ping times out, the server will be regarded as down by the Sentinel. However, this is only a subjective view of one Sentinel, namely subjectively down (+sdown in Sentinel events). In this example, Sentinel-1 marked the master as +sdown.
- To prevent false alarms, the Sentinel that marked ...