Hi, Igniters. Does anyone know or remember the reason of the additional
'backward ping' when some node loses the connection to its next in the
ring? The ticket is [1]. Consider: 1) There is a cluster with nodes 'A',
'B, 'C' 2) 'A' fails to negotiate with B by the discovery, tries to
establish a new permanent connection to 'C' and claims that 'B' is
failed. 3) 'C' may ping 'B' first instead of accepting the connection.
Do we really need the step #3? Some 1 node is most likely going to leave
cluster anyway: 'A' or 'B'. Why not just to recover the cluster
connection ASAP? Why is this ping considered more important than real
lost discovery traffic?
This extra node checking is now just a socket opening, not even a simple
write/read to the socket. Such an approach can lead to false positive
result of the node check and alive node 'A' would leave the cluster. The
ping might be more complex, but I doubt we need this check at all. [1]
https://issues.apache.org/jira/browse/IGNITE-7163
