A riak-nodejs-client user has noted a scenario in which one node has 2x the network traffic of the other nodes in the cluster. My initial theory is that a slow node is causing commands to "pile up" on this node.
More information needed:
riak-debug output from slow and normal node
A description of the commands being sent to the cluster - % that are read, % write, and if there are keys more frequently accessed than others
Proposed solution: in the default node manager, take the current # of executing commands on a node into account when selecting a node for the next command. Try to keep the execution # the same or close to it for all nodes.
A
riak-nodejs-client
user has noted a scenario in which one node has 2x the network traffic of the other nodes in the cluster. My initial theory is that a slow node is causing commands to "pile up" on this node.More information needed:
riak-debug
output from slow and normal nodeProposed solution: in the default node manager, take the current # of executing commands on a node into account when selecting a node for the next command. Try to keep the execution # the same or close to it for all nodes.