Hi! It was suggested from the Project Volemort forums that I post my question here:
I have a Project Voldemort 4 node BDB cluster that is at about ~80GB per node. I'm in the process of testing a migration of the cluster and its data to new servers. The current servers are Solaris, the new servers are Linux. I'm doing a 1 for 1 copy of the node data, keeping everything identical, except the server names in cluster.xml. It looks like this:
Old_Node1 --> New_Node1
Old_Node2 --> New_Node2
Old_Node3 --> New_Node3
Old_Node4 --> New_Node4
After copying the data over and starting up the new nodes, everything works great. But, I did notice that each node exhibits high CPU usage for a few hours (mind you the cluster is not servicing requests during this time), then eventually levels off to minimal CPU usage (which is typical of an idle cluster). An strace shows continual reading of the .jdb files during this time. I changed the log to DEBUG hoping to see some detailed information on what is happening, but noticed nothing out of the ordinary.
Could someone please provide some detail on what is happening during this process? Is it related to starting the cluster on new, different nodes? Some type of integrity check, etc?