Troubleshooting and FAQs › Troubleshooting the Manager › Repairing Job Records Containing Stale Cluster Aliases › Using NSCHED$ALIAS_GROUPS to repair cluster aliases

Using NSCHED$ALIAS_GROUPS to repair cluster aliases

Notes:

• The word "name" is used to refer to either a cluster alias name or a single node name.
• Only define NSCHED$ALIAS_GROUPS if cluster alias changes are made or planned. The product works more efficiently if the logical name is not defined.

Define the logical name on remote clusters and/or nodes which have dependency relations with the cluster where the alias is being changed. It does not need to be defined in the cluster where alias name gets changed. In other words, the logical name should not contain any local cluster alias or node names. Dependency records are updated when a job attempts to run and checks its dependencies and when "dependency job completed” notifications are sent.

If a cluster alias has been changed and the old alias is not accessible in the network, define the logical name equivalence string to include both old and new alias in order for Job Management Manager to resolve the dependencies.

If a cluster uses multiple aliases, define the logical name equivalence string to contain all of them.

Individual cluster node names are also allowed in the equivalence name. However it only makes sense to include a node name if a job dependency was set using that node name.

The logical name can be defined as a search list where each equivalence name represents comma separated names for one cluster. This way a single logical name can be used to resolve names for multiple clusters.

WARNING: Each equivalence name may only contain names for the same cluster. Putting names for different clusters into the same equivalence name is invalid and might lead to unpredicted results.

When the logical name is properly defined, Job Management Manager resolves conflicts between names in job records and actual names in the network. It then updates these names in the job database automatically.

If a name that is defined in the job record cannot be connected to, Job Management Manager attempts to connect to alternate names defined in the logical name until a successful connection is made, after which it updates the job record with the successfully connected name.

If a dependency job completion notification is received from a node whose name differs from the one defined in the job record (and both names are defined in the same equivalence name), Job Management Manager updates the dependency record using the name the notification came from.

The recommended order of names in each equivalence name is as follows:

1. Old (inactive) name
2. Active cluster alias
3. Individual node name

The logical name can be defined in the LNM$SYSCLUSTER_TABLE to be used by all cluster nodes. Alternately, it can be defined in LNM$SYSTEM_TABLE on nodes which need job dependency node name resolution.

A single alias or node name length cannot exceed 6 characters.