Dear doc guys, this feature [1][2] deserves an article in the mod_cluster documentation, somewhere near that "ProxyPassMatch/ProxyPass" note in the current version. [1] https://issues.jboss.org/browse/EAP6-172 [2] https://bugzilla.redhat.com/show_bug.cgi?id=987259 Yesterday, on the 9th of March, we received information from the customer that shall enable development to engineer a configuration that will go into the documentation. On the actual contents of the documentation, please, consult Jean-Frederic Clere <jclere>. Questions regarding planning, time and the status of ACKS/NACKS on the [1][2] will be answered by John Doyle <jdoyle>. THX
A hotstandby node in the mod_cluster logic is a kind of last ressort node: when all other nodes of the cluster are down/in error the requests and the sessions will be routed to that node. To configure a hotstandby node just configure the node with factor=0 like: +++ <subsystem xmlns="urn:jboss:domain:modcluster:1.2"> <mod-cluster-config advertise-socket="modcluster" connector="ajp"> <simple-load-provider factor="0"/> </mod-cluster-config> </subsystem> +++ In mod_cluster-manager the node will be displayed with status OK and load=-1 Scenario with 3 node: A factor=10 B factor=10 C facctor=0 Normal behaviour load will be balanced between A and B, if one of the 2 get in error or stopped all requests will go to the other one, if both are in error or stop all request will go to C.
I find several things being wrong in the documentation. - naming conventions - it's Load: 0 not Load: -1, see [2] - swapping terms "Load" and "factor" - somewhat misleading wording Please, see the new version of the documentation text: == BEGIN == A hot-standby node in the mod_cluster logic is the last resort node to which all requests are routed if all other nodes are down. This is similar to the hot-standby logic in mod_proxy. To configure a hot-standby node, replace the dynamic-load-provider in mod_cluster subsystem with a simple-load-provider with factor set to 0, e.g.: <subsystem xmlns="urn:jboss:domain:modcluster:1.2"> <mod-cluster-config advertise-socket="modcluster" connector="ajp"> - <dynamic-load-provider> - <load-metric type="busyness"/> - </dynamic-load-provider> + <simple-load-provider factor="0"/> </mod-cluster-config> </subsystem> In mod_cluster-manager console [See Development_Guide, 9.5. Apache mod_cluster-manager Application][1], the node is displayed with OK status and Load: 0. For instance, if there are three nodes: Node A, Load: 10 Node B, Load: 10 Node C, Load: 0 The load will be ballanced between nodes A and B, unless they are both unavailable. In such case, node C will take the load. == END == [1] http://documentation-devel.engineering.redhat.com/site/documentation/en-US/JBoss_Enterprise_Application_Platform/6.3/html-single/Development_Guide/index.html#sect-Apache_mod_cluster-manager_Application [2] https://github.com/modcluster/mod_cluster/pull/95
I'm sorry, this is wrong: the XML configuration snippet is wrong (both providers are there at the same time), please, read my comment 5 , it says "replace", not "add". These + and - prefixes are well known diff syntax, but you might want to use your own specific formatting or convention. The only point is to show both states: before and after the change. Furthermore, why isn't the text "refer Apache mod_cluster-manager Application section in the JBoss Enterprise Application Platform Development Guide" an actual link? These are small bugs, so we are almost there :-) Thanks
Topic modified/updated: 18.3.4. mod_cluster Configuration on httpd Summary of changes/updates: - Included information as per comment 5 after the Table 18.2. mod_cluster Derivatives Link to topic: http://documentation-devel.engineering.redhat.com/site/documentation/en-US/JBoss_Enterprise_Application_Platform/6.3/html-single/Administration_and_Configuration_Guide/index.html#mod_cluster_Configuration_on_httpd Build ID: JBoss_Enterprise_Application_Platform-Administration_and_Configuration_Guide-6.3-web-en-US-6.3.0-28.el6eng
It's O.K. now. THX.