Show
added a comment - scenario:
there are two open UI sessions
1- user#1 tries to remove couple of nodes and hits rebalance
2- rebalance is stopped or it fails
3- then the user will want to "continue" the rebalance operation
4- if the user clicks on rebalance from a different web UI then UI wil not remember the lsit of ejected_nodes and it will add everthong to knownNodes
this is a big problem in large clusters with millions of items because
1- user has no idea that the previous attempt was rebalance out and re-rebalancing will move millions of items
there are two open UI sessions
1- user#1 tries to remove couple of nodes and hits rebalance
2- rebalance is stopped or it fails
3- then the user will want to "continue" the rebalance operation
4- if the user clicks on rebalance from a different web UI then UI wil not remember the lsit of ejected_nodes and it will add everthong to knownNodes
this is a big problem in large clusters with millions of items because
1- user has no idea that the previous attempt was rebalance out and re-rebalancing will move millions of items