Version 2.1 broken cluster?

[rkozlo]

Hi there,
testing latest stable version 2.1. Trying to set cluster but i'm stuck. Everything seems to be correct but for some reasons i have an error
CLUSTERX SETNODES "ae23b94ccaf714337e4ce5ba99ef3dc257e300df 10.32.68.251 6666 master - 0-5460 32d332da761f44df7959e5887b6b94cb4667d781 10.32.68.250 6666 master - 5461-10992 862a51f8b6294a4b0729a7c5c929bfd67068d962 10.32.68.249 6666 master - 10993-16383" 1
and error
(error) Slot distribution is overlapped
Changing slots doesn't change a thing.
Hit this command with only one node passes without problem. Is it some bug or unfinished code? DId everything basing on doc. Didn't find any related issues.

[git-hulk]

You can use the redis-cli cluster slots to have a look at current slots in node. Current implementation have a bug that it can't set the slot to node even the current owner is itself.

[git-hulk]

Thanks @rkozlo

For the first question, should add a new line between nodes:

clusterx setnodes "kvrockskvrockskvrockskvrockskvrocksnode1 10.32.68.251 6666 master - 0-5460 \n kvrockskvrockskvrockskvrockskvrocksnode2 10.32.68.250 6667 master - 5461-10992 \n kvrockskvrockskvrockskvrockskvrocksnode3 10.32.68.249 6666 master - 10993-16383" 1

And for the second question, it works well since the version is the same as the previous command and won't check the node's format, so didn't return the error.

[rkozlo]

Ok, it works. One question and i'm out ;)

127.0.0.1:6666> cluster help
(error) ERR CLUSTER command, CLUSTER INFO|NODES|SLOTS|KEYSLOT
127.0.0.1:6666> clusterx help
(error) ERR CLUSTERX command, CLUSTERX VERSION|SETNODEID|SETNODES|SETSLOT|MIGRATE

only those options for cluster and clusterx are supported? No failover?

[git-hulk]

Yes, the failover for the Kvrocks cluster is to change the cluster topology. It's expected to do outside the kvrocks like Kvrocks Controller(working in progress now)

[rkozlo]

@git-hulk thanks for help, you can close this topic, because i can't

[git-hulk]

@rkozlo Thanks for your feedback too. Sure, let's lock this discussion.