If it is not possible to perform a warm start for an abnormally terminated node application at the node's own node computer in reasonable time and also no virtual host has been defined then a node recovery can be performed for this node on another node in the UTM cluster in order to avoid impairing the performance of the running UTM cluster application.
Prerequisites for the use of node recovery
Node recovery requires the presence of SYSLOG files with node-specific names that can be accessed throughout the cluster.
You can dynamically generate the start parameter file with the required node name for the NODE-TO-RECOVER start parameter.
Alternatively, you can provide, for each node in the cluster, a previously set up start parameter file for node recovery that can be accessed throughout the cluster.
If you have departed from the recommendations and installed UTM or other runtime components under different paths on the individual cluster nodes and if this code is loaded from shared objects, then you should note the following:
For it to be possible to call node recovery, the application must have been appropriately linked for this type of use.
In addition, you must set the environment variables $LD_LIBRARY_PATH and, if necessary, $LD_LIBRARY_PATH64 to the locally accessible paths, i.e. as they are at the start of the local node application.
For more detailed information, see the sections "Installing the UTM runtime components for Unix and Linux systems" and "Installing further runtime components for Unix and Linux systems".
Starting node recovery
Node recovery is controlled via the start parameters listed below.
NODE-TO-RECOVER
selects a node in the UTM cluster application for which node recovery is to be performed.
RESET-PTC
specifies whether or not transactions in the PTC state are to be reset on node recovery.
For a more detailed description of these start parameters, see section "Start parameters for openUTM".
Calling utmmain for node recovery
| Start the program utmmain as a background process (see "Starting a UTM application on Unix and Linux systems" for Unix and Linux systems, "Starting a UTM application on Windows systems" for Windows systems) in order to start node recovery. When doing this, specify the filebase name of the node application for which node recovery is to be performed as the first argument and start node recovery in this filebase directory. CAUTION! The start procedure for starting the node application that is to perform node recovery must not contain any commands that have an effect on node applications running in parallel on this node computer. This includes, for example, a call of the utility program kdcrem prior to the start of utmmain. |
Messages
When node recovery is started, the message K192 is sent to stdout and stderr. This message logs the values of the start parameters NODE-TO-RECOVER and RESET-PTC together with the current computer name.
A K193 message is output for every detected transaction with the PTC state, irrespective of the value of the RESET-PTC parameter.
A K160 message is output for every transaction that is reset.
At the end of node recovery, a K194 message is output which indicates the number of GSSB and ULS areas still locked by this node.