Binbin
`n-1` is a hard-coded number, if the primary got more replicas, it will still got the error here. One way is to update num_replicas when the replica is disconnected (we...
i am guessing we can't simply ignore the NOREPLICAS error here? otherwise we will lost the CLUSTER SETSLOT command in the source node view. (though it may be harmless, the...
force-push for the DCO.
@PingXie thanks for the review! i think i took care of all the comments, please take another look.
> did we have a test failure here, or should we introduce new test for this case? emmm, i don't think we have a test failure here, we don't have...
> It's not a breaking change? The behavior change is that if the sync fails, the old data is still there, instead of deleted. It seems better. Nobody relies on...
i am going to merge this one, i don't think it is a breaking change, does any of you have other concerns?
i think it should wait the online replica, but you also got a good point here, so in this case, you are suggesting user should fix the offline replica first...
we are struck in this must or not situation. I just thought a good point about why we should use online replica. Migrations is some kind of admin operation, if...
It occurred to me that maybe we also want to cancel server.rdb_bgsave_scheduled? Noted it will also cancel the replication child process. Also does AOFRW need the same cancel? seem to...