It’s pretty certain that setting a maintenance reservation instead of draining nodes in advance of the reservation would at least change the message. I’m not sure if it will make more sense or not, but I’d think it might.
> On Jul 5, 2016, at 10:23 AM, E.M. Dragowsky <dragow...@case.edu> wrote: > > Greetings -- > > A few users have experienced jobs pending due to the reason > ReqNodeNotAvail(Unavailable:<nodename1>,<nodename2>,...) > We have determined that the jobs in fact were pending due to asking for > TimeLimit > "time remaining before maintenace shutdown" -- managed by making > a global reservation on all nodes, all partitions. > > This reason code is not very helpful in understanding the reason for the jobs > being pending. Resubmitting the jobs with appropriate TimeLimit allows the > jobs to run immediately. The jobs were therefore pending due to "excessive > time requested". > > To prehaps help knowledgeable developers understand why the Reason > ReqNodeNotAvail appeared, I note that the nodes listed in the above are > actually being drained in advance of updates. Happy to provide further > information as needed. -- ____ || \\UTGERS, |---------------------------*O*--------------------------- ||_// the State | Ryan Novosielski - novos...@rutgers.edu || \\ University | Sr. Technologist - 973/972.0922 (2x0922) ~*~ RBHS Campus || \\ of NJ | Office of Advanced Research Computing - MSB C630, Newark `'
Description: Message signed with OpenPGP using GPGMail