Redis-sentinel: not able to establish correct connection about redis HOT 8 CLOSED

I made a fix for the communication issues.
However I was not able to get the feature to work as is. I ran into several timing issues. I tried some fixes, but still had problems. Looking in the original redis repo I see many changes to the sentinel code after 2.6.12. I think this code is not yet stable, so I will wait before coming back to this.

from redis.

Thanks for looking into this issue. I quickly created a copy the start_all script to run on OSX/Linux, and ran it against the latest 2.6 version (installed via brew on my mac). That seemed to work; at least, establishing the sentinel connections, and registering the available slaves.

I'll check back; in the meantime, is there some other recommendation for HA? I was chatting with some of the SignalR guys in explaining our original scenario (HA for SignalR), and they suggested looking at the Bootsleeve connection factory (which they are wrapping for the Redis backplane).

from redis.

why should call aeWinReceiveDone after redisAsyncHandleRead in redisAeReadEvent?
Isn't that redisAsyncHandleRead always call aeWinReceiveDone to add a new receive event?

from redis.

The aeWinReceiveDone needs to be called after reading the available data. It looks like redisAsyncHandleRead does it before reading the data.

from redis.

The async connect module used by sentinel has been updated to work properly in Windows. With this change the sentinel code seems to be working.

from redis.

Very cool, Henry. I'll check it out - first on my referenced test repository, then in our test environment.

from redis.

Does anyone have sentinel working correctly in windows? When I run sentinel mode it seems to connect, but it always thinks the master is down (when it's not) and when I "sentinel masters" it says that it has no other sentinels (there are 2 others) and in fact no slaves. Redis itself is running fine. My master does have a slave and that is working fine. I'm using 1.6.8-pre2 64 bit

from redis.

I just checked in commit ff367b5 into the 2.8.4_msopen branch that fixes the scenario mentioned in the first post of this thread. Please let me know if this works for you.

from redis.