[chart/redis-ha] Using a more smart proxy

[mhkarimi1383]

Is your feature request related to a problem? Please describe.
When master changes HAProxy has a delay to understand that changes and also it will not shutdown clients, It will cause race condition some times

Describe the solution you'd like
In @ParminCloud We have made a project https://github.com/ParminCloud/haproxy-redis-sentinel It will use HAProxy Runtime API to handle that change live using pub/sub

Also we have made a Helm Chart for that

Describe alternatives you've considered
I am currently installing this chart without haproxy enabled, Then installed our chart separately and connected them together, And results are awesome

[DandyDeveloper]

@mhkarimi1383 What's the latency difference here? Because in my case, Redis seems to lose its master so infrequently, I don't think I've ever had an impact from the HAProxy failover in production scenarios.

How many writes do you have happening per second? I want to understand the use case a bit better to justify the change .

[mhkarimi1383]

@mhkarimi1383 What's the latency difference here? Because in my case, Redis seems to lose its master so infrequently, I don't think I've ever had an impact from the HAProxy failover in production scenarios.

How many writes do you have happening per second? I want to understand the use case a bit better to justify the change .

In some scenarios where the network becomes unstable Sentinel changes master while we are under the load and some of critical jobs run multiple times (e.g. GitLab CI Jobs)
(I know that master should not change frequently in a normal environment, but we can protect our application)

We have also forked this chart and separated Sentinels from Servers

This is more stable right now ;)