Jay
Now more logs are added, we should get more insight once it's reproduced.
This is somewhat expected to me. When a node is applying snapshot, then it should be considered in an unhealthy state. If another peer is isolated from the group, then...
> ..the kv client could not get the response in time and the request sender will keep retrying.. I don't get it. For example, the new leader returns error TooBusy,...
> v1 and v2 should produce same results. Do you have further interest to reduce the usage of v1? :) I have the same confusion. For example, the only difference...
I think the problem is due to the race between (1) replicating peer that has small ID, (2) split and (3) replicating peer that has larger ID. During being destroyed,...
I can see there are several PRs that tries to update dependencies. How about also include those changes in this PR? https://github.com/tikv/tikv/pulls/app%2Fdependabot
It's not about free space, but the util/latency of disk.
Does such use case exist? Is it a bug in real life or just an unexpected behavior for possible usage?