Monitoring - The fix for the issue was surprisingly easy to deploy. Users should no longer encounter this issue. Please let us know at chtc@cs.wisc.edu if you do!
Jan 09, 2026 - 15:27 CST
Update - Slight correction: without "max_idle", you can submit up to 10,000 jobs per submission and you can have up to 50,000 jobs total in the queue. ("Done" jobs generally do not count against the total in the queue.)
Jan 09, 2026 - 14:47 CST
Identified - We've identified an issue where submit files that have "max_idle" (and related keywords) may not create jobs when submitted with "condor_submit". We have narrowed down the source of the bug, but do not have a fix at this time (and probably not before the weekend).
If you will have less than 10,000 jobs in the queue, we recommend removing this option from your submit file. If you will more than 10,000 jobs in the queue, you may want to consider removing this option and manually submitting your jobs in batches such that the total number jobs stays under the 10,000 job limit.
We appreciate your patience while we work to deploy a fix for this issue.
Jan 09, 2026 - 13:57 CST
Update - The affected GPU has been removed from the available pool for testing. Unaffected GPUs on the machine are available for use. We are still investigating.
Jan 09, 2026 - 10:35 CST
Investigating - Some jobs landing on xhuanggpu4001 may fail with the message, "uncorrectable ECC error encountered". We have narrowed the issue to a specific GPU. We plan to bring the affected GPU offline for further investigation and fixes, which may require a brief downtime. The machine will be brought back up with the unaffected GPUs available for use.
Jan 07, 2026 - 13:23 CST