Closed sunheehnus closed 9 years ago
AddJob or GotJob message may be lost from socket. We should resend them to avoid a client being blocked forever.
This situation may happen in small scale cluster.
node1: send AddJob (using clusterReplicateJob and node2 is already added to node1->nodes_delivered) node2: get AddJob ,send GotJob and GotJob is lost.
clusterReplicateJob
node1->nodes_delivered
client who addjob to node1 will be blocking until timeout.
:-)
Oh, sorry, after a second thinking. We are using TCP to transfer data which will make it reliable. This patch will only cause flooding in network. >_< Closing.
AddJob or GotJob message may be lost from socket. We should resend them to avoid a client being blocked forever.
This situation may happen in small scale cluster.
2 NODES CLUSTER
node1: send AddJob (using
clusterReplicateJob
and node2 is already added tonode1->nodes_delivered
) node2: get AddJob ,send GotJob and GotJob is lost.client who addjob to node1 will be blocking until timeout.
:-)