Open fcasson opened 4 years ago
Differentiating between impossible requests and requests requiring queuing depends on https://github.com/prominence-eosc/imc/issues/37
The alternative mentioned above depends on https://github.com/prominence-eosc/imc/issues/59
With this commit https://github.com/prominence-eosc/prominence/commit/d381bb97c90136a8737687088134c7b71468ee26 and the commits mentioned in https://github.com/prominence-eosc/imc/issues/37#issuecomment-589433555 there are now two possiblities for failed jobs related to resources:
No matching resources
: no resources exist which match the jobs requirementsNo matching resources currently available
: no resources are currently available to run the jobIf a job requests resources which do not exist and maximumTimeInQueue
is non-zero, the job will quickly fail, because remaining in the queue will be of no benefit. However, if resources do exist but they are busy, the job will remain in the queue.
Will add a prominence explain
(or something similar) which will say why an idle job is not yet running.
If a job fails, or is "queued", due to an unprovisionable resource request, it is useful for users to know which part of the resource request is limiting. It is also helpful to differentiate impossible requests (resources do not exist) from requests which require queuing (resources are in use).
An alternative is just provide information on the currently existing / used / available resources in various projects (including opportunistic (c.f. #118) options), so users can tailor their resource request to match the availability (most batch queue systems provide ways to do this).