Ping Xie

@madolson, this initial slot metrics set make sense to me. What do you think about the importing/migrating states for a slot? For input, I like both the `SLOTS` token and...

> To handle the first component, we plan on introducing a size parameter of type size_t for data-structures that are not contiguous in-memory (thus, we can’t track it’s memory through...

Thanks for the detailed write-up, @kyle-yh-kim. With your explanation, the biggest downside to me with the TLS idea is the challenge to filter out temporary memory allocations whose life cycle...

> `CLUSTER MIGRATESLOTS [NODE ] [SHARD ] ' Can we add the slot range syntax sugar similar to [`ADDSLOTSRANGE`](https://redis.io/commands/cluster-addslotsrange/) for consistency? How do you plan to support auth? Are you...

@zuiderkwast I think you are essentially suggesting 2PC in the target shard as well, either hanging off of the 2PC between source and target primaries or part of the same...

Right, the one benefit with this proposal (in its current "forking" form) is that clients no longer need to deal with redirection during the slot migration. Design wise, the cluster...

@madolson I see strong value in getting this change rolled out together with `CLUSTER SHARDS` as it helps complete the shard scenario end 2 end. Additionally, it has externally visible...

> @PingXie I see a lot of your points, and I agree with most of them. I have two concerns. Let's focus on the implementation, if it's ready we can...

@madolson I pushed another commit to make `CLUSTER SHARDS` work better with failed primaries and this should make the shard id support complete. Below are the externally visible changes (hence...

> Regarding the form `ip:port@cport,shard,hostname` (which is already a pretty strange format without shard), I have two incompatible suggestions: > > 1. Since every part so far uses a different...