I think we both understand. If you can provide more details about your workload and environment, that will be useful. Thanks again!
I have RHEL5, 6, 7, SLES10, 11, 12, UBNT1404, 1804, AIX and more…
Sometimes the user want to submit the job for some kind of OS, you can see in my previous posts.
Right now I have the huge string for OS version, with all possible combinations assigned like the static resource on each node. But I want to assign just only one resource and cluster understand “or” logic.
Tell you more) its not related just only for osv. What about different logic like “more”? For example memory more than 20G. Find the host with memory more than 20G.
Any updates? Trying to resolve the issue. Please, help.
@Ve0 It seems what you want for your use case is a way to filter out nodes that you want workload manager to consider. Which is what is being worked out as " https://pbspro.atlassian.net/browse/PP-507 ".
While it is being worked out, a way to make this work for you is to use string_array resources. In case of memory as well you can have a resource “res_mem” set like “osv” (mentioned in previous examples) and request them in your job request.
Now, the question I have is that if you know a job needs high memory nodes then it must also be requesting higher memory right? then, in that case, why do you really need the classification of high memory nodes?
Hi – it sounds like you may be trying to migrate from another product to PBS Pro, and you are trying to exactly replicate the submission interfaces for the users. In my experience, many folks have migrated successfully to PBS Pro (from all the other workload managers), but, since all the products provide different models of workload management, it sometimes requires changes in approach. This is why I was asking for more higher-level information on your system, hardware, user base, workload, performance goals, etc – one or two levels higher than “a jobs needs X or Y resources”.
Without more higher level information, the best way to handle your use case with the existing PBS Pro v18 is to use the string array approach. In your case (again as understood from the low-level), it will be very “wordy” with long string arrays. There are other approaches, but without a lot more site-specific/workload-specific details, it’s hard to zero in on anything.
It sounds like your use case is similar to what is being proposed in PP-506/507, but that’s not currently being worked on (more than design discussions) right now. I’m not sure how important this is for your site, but Altair does offer professional/development services. If your organization is willing to fund that development, that’s an option – let me know.
You’re right, we try to find the best solution to migrate. Trying the BPS. But PBS has a lot of limitations. Even the hooks do not save us from it.
Yes, trying to find something like PP-506/507 in PBS, but no luck. This is the showstopper for us.