HOST SPECIFIC MONITOR SCRIPT FOR NETWORKED COMPUTER CLUSTERS

A monitoring subroutine is embedded within a package that provides no other client services other than the monitoring functionality itself. Because of this, the monitoring package can be shut down at any time without affecting any other system services or packages. The inventive concept is realized by providing one monitor package for each type of package running on a server; the one monitor package running on the same server. Also, this invention provides a communication mechanism between the monitoring subroutine and the packages whose services are being monitored. This is necessary to provide the monitoring subroutine a real time view of the resources being provided by all of the packages as a whole. To accomplish this, a state table is provided for all packages of the same type running or pending on a given node and similar monitor state tables are provided for other nodes. In this implementation, the standard packages just make one-time changes to the system resources. They make a change in the state table when they start, and undo it when they stop; once again, adjusting the state table, so a monitor program knows what the current state of resources, packages and nodes should be. The monitor subroutine embedded in a monitoring package loops indefinitely and constantly checks the real state of resources against the state table. Now, if any of the system resources fail then the monitor package restarts the failed system resource and restores it to its “last known use” state without stopping and re-starting any of the packages on that server that were using that system resource.