Safely change configuration files

It is important and highly recommended to configure EGO through the cluster management console, not directly within the various configuration files.

Be aware of cause-effect relationships that exist in EGO between files. For example, manual changes made to the ego.conf, ResourceGroup.xml or ConsumerTrees.xml files may potentially affect other EGO configuration settings.

Furthermore, to achieve a specific goal, a whole set of related parameters must often be changed or tuned collectively, some in obscure directory locations. If not done correctly, the output of manually updated configuration files may not produce the expected behavior.

Finally, with certain configuration settings, important validations triggered through the cluster management console must be conducted by EGO. By using the cluster management console, legitimate and allowable parameter settings are ensured.

Examples of potential problems caused by manually configuring files:
  • If you manually configure a consumer policy within the DistributionTree sections of the ConsumerTrees.xml file without adding a corresponding instance in the ConsumerHierarchy section, then EGO does not recognize the newly added consumer.
  • If you manually configure borrow and lend plans without giving a full consumer path (including the consumer tree name), then EGO may ignore them.
  • If you set an invalid time window (not covering a 0 to 24 hour time period), then resource plans do not behave as expected.
  • If you allot an unbalanced ownership, (such as reconfiguring a leaf consumer’s ownership value without changing the value at the branch level), then resource plans are not be effective and workload units do not run as expected.
  • If you delete a consumer manually without checking whether it owns allocations or is currently running activities, you can seriously affect your resource plan and running workload units, among other things.