The clearcase plugin always uses the current computer for resolution of build variables rather than the Computer associated with the last build, even when polling on the last build's Computer. This causes polling to fail for us. We allow our jobs to roam freely in the cluster. But based on the code, even if the job was tied to a particular node and the node was not the master node, AND the view-name depends upon node-specific configuration, the polling fails.

To let our jobs roam freely in the cluster, we introduced a node-specific attribute to the view name. Our views are all named ${SHORT_COMPUTERNAME}_${JOB_NAME}. SHORT_COMPUTERNAME is a parameter uniquely defined on each Node. This is so that clearcase doesn't complain when there are 2 views for the job in the cluster.

hudson: 1.339
clearcase: 1.1