Open ben-yacine opened 4 years ago
If some one can help please. I'm still stuck in this step. Thanks in advance.
Nicely described. I am also facing same. On Power off of myster server(node), failover not happening. We can see new nominated master in sentinels.
net
package doesn't trigger any event of error on that situation, that's the reason failover doesnot work. That means, current implementation of failover only works if we gracefully shutdown of redis server.
Has anyone found a manual way to failover manually? I have noticed the same issue.
I found the same problem. If I kill a pod with Redis master node, ioredis doesn't notice a new master elected by other sentinels.
I have the same problem. With current implementation sentinel is useless.
Hello,
I have the following config of my redis stack. One master redis and one slave connected to this master. I also create a sentinel instance to monitor the redis master. I can connect to sentinel with ioredis and everything works fine.
My problem is when i execute the command
DEBUG SLEEP 80
to cause the failover, sentinel detect that the master is down and promote the slave as master. But after that, sentinel will try to make the previous master as slave but it doesn't succeed because the master still down and we have+SDOWN
log in sentinel part. During this period, ioredis will not redirect the request coming to the new master until sentinel make the master as slave. so ioredis will connect again to the new master and the requests will be treated.What i want is if we can make the requests redirected to the new master even if the last one is not yet promoted as slave. this is the high availability that we want to have from using sentinel.
Any idea ?
Thanks in advance.