Paragrf commented on issue #3385:
URL: https://github.com/apache/kvrocks/issues/3385#issuecomment-4037272994

   > * Redis is supposed to persist the fail status into nodes.conf, but it 
doesn't appear to be doing so right now. Do we need do it?
   > * Are there specific fields in CLUSTER INFO (such as, if you set 
master,fail we need change cluster_state,  cluster_slots_ok, 
cluster_slots_fail) that should also be updated to reflect this?
   
   @jihuayu Thank you for your suggestions.
   
   First, Redis nodes are autonomous, and their persisted nodes.conf files 
ensure that a crashed node can rejoin the cluster without losing its topology. 
In contrast, Kvrocks employs a centralized management scheme where an external 
controller is responsible for reconfiguring the topology, making the 
persistence of old topology unnecessary. Furthermore, persisting such data 
could easily lead to a state misalignment between the server nodes and the 
management layer.
   
   Second, The original intent of this modification was to handle client 
traffic removal (deregistration) specifically for replica node downtime, which 
theoretically should not involve changes to CLUSTER INFO. While this PR does 
introduce the capability to modify master node states, as long as the 
controller does not explicitly trigger those changes, the issue you're 
concerned about should not arise.
   
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to