Description of problem: In low-latency, a job that is rejected will be released for another startd to grab however there is nothing preventing the node that just rejected the work from attempting to run it again. Sometimes immediately after releasing the work if the timing lines up. A means to address this is to have rejected work removed from the work queue, modified to add/modify an attribute denoting node names the job should not run on, and re-submit. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Reopen when feature is required.