|
|
|
Two nodes are added to the ALPHAS cluster. New nodes are IA64 architecture, therefore it is decided that they use a different cluster alias, IA64S. Job Management Manager has been installed on the new nodes and all four cluster nodes use the same job database. The jobs ALP1, ALP2, ALP3 and ALP4 are allowed to run on any node in the cluster.
Implications: When jobs ALP3 and ALP4 run and complete on one of new IA64 nodes, their dependents PLANET::PL1 and PRD1/PRD2::PROD3 get stuck in DepWait state and never run. Additionally, if the two Alpha nodes ALPHA1 and ALPHA2 are down but the IA64 nodes are running, jobs ALP1 and ALP2 will not run when their dependencies PLANET::E1 and PRD1::PROD1 are complete.
Solution:
Define the logical name NSCHED$ALIAS_GROUPS as follows on both the PLANET and PRD1/PRD2 clusters:
$ DEFINE/TABLE=LNM$SYSCLUSTER_TABLE NSCHED$ALIAS_GROUPS "ALPHAS, IA64S"
Results after defining the logical name:
When jobs ALP3 and ALP4 run and complete on one of new IA64 nodes, their dependent jobs PLANET::PL1 and PRD1/PRD2::PROD3 start as expected and their dependency records are changed as follows:
Job PLANET::PL1 depends on IA64S::ALP3
Job PRD1/PRD2::PROD3 depends on IA64S::ALP4
If nodes ALPHA1 and ALPHA2 are down, jobs ALP1 and ALP2 start on an IA64 node and their dependency records are changed as follows:
Job IA64S::ALP1 is dependent on job PLANET::E1
Job IA64S::ALP2 is dependent on job PRD1::PROD1
| Copyright © 2012 CA. All rights reserved. | Tell Technical Publications how we can improve this information |