Cluster Monitor Reference

The cluster monitor is useful in environments not running HP OpenView ClusterView. The cluster monitor reports information on the status of the cluster to which the local node belongs. The resources monitored are:

Cluster status, (/cluster/status/clusterName), a summary of the state of all nodes in the cluster name.
Local node status, (/cluster/localNode/status/clusterName), the status of a given node in a cluster; if the node is part of more than one cluster, status is given for each cluster to which that node belongs.
Package status, (/cluster/package/status/packageName), the status of an MC/ServiceGuard package on a node/cluster.

To fix any problems detected by the cluster monitor, refer to Managing MC/Service Guard.

Figure 3-1 Cluster Monitor Resource Class Hierarchy

Figure 3-1 “Cluster Monitor Resource Class Hierarchy” shows the cluster monitor class hierarchy.

Items in boxes are resource instances that can be monitored.
Items in italics change depending on the names of the clusters and packages on the system.

Cluster Status

The cluster status is the status of the MC/ServiceGuard cluster to which this node belongs.The status is from the perspective of the node for which the request was created.

The MIB variable hpmcClusterState, which is part of the hp-mcCluster MIB, provides the cluster status information to the monitor.

The cmviewcl -v command displays detailed information about the current status of the cluster and packages on the cluster.

Table 3-1 Interpreting Custer Status

Resource Name		/cluster/status/`clusterName`
Condition	Value	Interpretation
UP	1	The node can access the cluster.
UNKNOWN	2	The node may be separated from other active cluster elements (for example the heartbeat LAN) and has insufficient information to tell if the cluster is accessible.
DOWN	3	The node cannot access the cluster.

You might request to be notified when the cluster is not up. You could then verify that the cluster was shut down intentionally.

The minimum polling interval for cluster status is 30 seconds. You may want a longer interval, especially if system performance is affected.

Node Status

The node status is the current status of a node relative to a particular cluster.

The MIB variable hpmcClusterState, which is part of the hp-mcCluster MIB, provides the node status information to the monitor.

The cmviewcl -v command displays detailed information about the current status of the cluster and packages on the cluster.

Table 3-2 Interpreting Node Status

Resource Name		/cluster/localNode/status/`clusterName`
Condition	Value	Interpretation
RUNNING	1	Node is accessible and operating normally.
INITIALIZING	2	Node's daemon has started, but is not ready to communicate with other nodes' daemons.
RECONFIGURING	3	Node is running protocols to make sure all other nodes agree to the new membership in the cluster.
INVALID	4	The cluster status may be DOWN.
HALTED	5	Node has been removed from the cluster, with the cmhaltnode command, for example.
FAILED	6	Node is no longer a member of an active cluster.

You might want to create a request that notifies you when the local node is not running. You can then verify that the node or MC/ServiceGuard was stopped intentionally.

The minimum polling interval for cluster status is 30 seconds. You may want a longer interval, especially if system performance is affected.

Package Status

The package status is the status of each package running on this node.

The MIB variable hpmcClusterState, which is part of the hp-mcCluster MIB, provides the package status information to the monitor.

The cmviewcl -v command displays detailed information about the current status of the cluster and packages on the cluster.

Table 3-3 Interpreting Package Status

Resource Name		/cluster/package/status/`packageName`
Condition	Value	Interpretation
UP	1	The package is running on the local node.
UNKNOWN	2	The package is not running on the local node, but may be running on another node in the cluster.
DOWN	3	The package is not running on any node in the cluster.

You might want to be notified when the value of any of the packages changes to UNKNOWN or DOWN, so you can verify that MC/ServiceGuard successfully migrated the package to another system.

You may see many packages with UNKNOWN status. This is because only the node running a package has complete status for a package. Other nodes often have inactive volume groups that make it impossible to have complete knowledge of package status. If a package is running on another node in the cluster, the current node may not have complete status on that package, and reports the condition UNKNOWN.

The minimum polling interval for cluster status is 30 seconds. You may want a longer interval, especially if system performance is affected.

Cluster Monitor Reference

Technical documentation

» Table of Contents

» Glossary

» Index

Cluster Status

Node Status

Package Status