Redis/Sentinel cluster failover results in "failover-abort-not-elected master"

Ignore me on the instance ID confusion, that's my fault - I forgot that the sentinels were voting for a leader among the sentinels, not voting for a new master among the candidates, so the three different IDs makes sense.

So, here's the real problem:


There's no slaves there that the sentinels have noticed, so they don't have a good node to make live during the vote.

Check the slaveof config on the instance, make sure that instance is running, and connect to it to check its info to verify that it's slaving. Once the sees it as a slave (in its info command), the sentinels will pick it up as a known slave and be able to fail over to it.