Solves problem where long standing idle tcp connections between clients and servers are covertly closed by intermediary firewalls.
…ying based on the remote metadata store. Cleans the code up a bit. Also relaxed serialization on the ClientStoreVerifier so that the actual get call occurs outside of the synchronized block. Alas, it should make no difference as there's only one thread performing verification ;)
…d as unavailable due to exceeding request timeouts Clean up of the logging to output a message detailing that the timeouts have been exceeded when marking the node as unavailable. Also cleaned up the logging to only output the node ID instead of the node ID *and* all of its partitions (which, in a production system is very verbose).
Synchronized store addition so concurrent editting/saving of configuration data doesn't occur: ConfigurationStorageEngine does not support that.
* Reissuing requests to stimulate caching on the server. Wrap the core KeyProvider in an implementation that pulls out previously issued key and reuse an entry from that set randomly. * Collect request times and store them in an array that is used to collect percentiles. Didn't add it to the "m" case as I'm not sure if that makes any sense.
- Created a tool to filter a request file for requests destined *only* for a specific node. - Added a -node option to RemoteTest to bootstrap only from a specified node rather than go to node 0 of the cluster.
…f assuming String.
… count exceptions as succesful for throughput counts.
…er to create metadata store based nodes on the fly. Also added a new unit test to ensure this.
- Default Serializer Factory now supports unversioned JsonTypeSerializers when store schema definition is "none". - Allow abstract store builder to use serializer factories other than Default Serializer Factory. - Fix read only utils to never return negative chunk numbers, even if Integer.MIN_VALUE is the key.
We clone the versions before doing read repair in the background.