Red Hat Bugzilla – Bug 997230
RFE: JMX Stats in local vs. clustered modes
Last modified: 2014-04-03 23:56:37 EDT
Details about the JMX Statistics available in JDGs local mode vs. clustered modes.
Scope: 1-2 topics specifying which statistics are available in local mode only, clustered mode only, or both.
Guide: Admin and Configuration Guide
Source: JBoss DataGrid questions
this is the info I was able to gather. Components (MBeans) that are available by default in clustered mode and are not in local mode. For individual components, I don't mention all attributes but only some of them. Not sure we need to mention every single attribute of every component in docs.
LocalTopologyManager (Controls the cache membership and state transfer)
attribute - rebalancingEnabled
JGroups protocols together with their attributes and operations:
FD_ALL (address: jboss.infinispan:type=protocol,cluster="clustered",protocol=FD_ALL)
DistributionManager (Component that handles distribution of content across a cluster):
RpcManager (Manages all remote calls to remote cache instances in the cluster.):
StateTransferManager (Component that handles state transfer)
Assigning to myself since I forgot to before emailing Martin.
Thanks for your response Martin!
Re: listing all attributes. We have the following Appendix to the Admin Guide which describes Attributes and Operations of MBeans in detail:
Is this level of detail not required for the components you mention above?
you're right, the level of detail should be probably the same. Looking at the Appendix, it seems most of the MBeans are already described well. Some of them just need a note about being available only in clustered mode.
One MBean that is not in the Appendix is LocalTopologyManager. So I'm providing more information. Feel free to rephrase it.
Description: Controls the cache membership and state transfer
description: If false, newly started nodes will not join the existing cluster nor the state will be transferred to them. If any of the current cluster members are stopped when rebalancing is disabled, the nodes will leave the cluster but the state won't be rebalanced among the remaining nodes. This will result in fewer copies than specified by the numOwners attribute until the rebalancing is enabled again.
I'm not sure whether we need to describe also the JGroups Channel and individual JGroups protocols. The JGroups Channel is available only in client-server mode (IMO) and the JGroups protocols might be different (different set of protocols). Maybe you can check with Tristan?