Describe the bug I am trying to perform a version upgrade without impacting the connected applications, but in several tests I have performed the old and new architecture end up mixing in an inexplicable way. Each architecture consists of: Old architecture 172.18.0.2:6380 -- master 172.18.0.3:6380 -- slave 172.18.0.2:26379 -- sentinel1 172.18.0.3:26379 -- sentinel2 172.18.0.4:26379 -- sentinel3 New architecture 172.19.0.2:6380 -- master 172.19.0.3:6380 -- slave 172.19.0.2:26379 -- sentinel1 172.19.0.3:26379 -- sentinel2 172.19.0.4:26379 -- sentinel3 I mean that both architectures have their configuration files and work perfectly independently, however the change I am trying to make is by changing the traffic through a NAT IP, which stops sending traffic to the old architecture and sends it to the new one.

But when that happens, the sentinels of the new architecture for some reason receive the information from the old architecture, get a new master, and the sentinels. So the configuration after the network change: 2 masters, 2 slaves, 6 sentinels. ó 1 master, 3 slaves, 6 sentinels The network change is the only change I make, but somehow the sentinels are messing with me by sharing configurations. I think I can solve this by bind, or by changing the ports, but I wondered if anyone else has gone through the same thing?

A short description of the bug. Sentinels share configurations and from 3 sentinels per architecture merge to be 6. To reproduce it is necessary to make a network hop with a NAT IP

Steps to reproduce the behavior and/or a minimal code sample. The strangest thing that happened is that one of the sentinels had the following configuration and yet it was filled with the information from the master and the other sentinels Redis [BUG] Sentinel configurations are mixed on 2 different architectures with same network segment and port

Expected behavior

That after the IP NAT change the configurations are not changed or mixed,

A description of what you expected to happen.

Additional information

Any additional information that is relevant to the problem.