Transfer compute jobs

To transfer compute jobs to PERSEUS, you can use the compute jobs API. This way, you can see all compute jobs related to a compute project within the project details view. Additionally, in case you use Andromeda, you allow your users to see their compute jobs as well.

You can transfer any number of jobs by using the endpoint /service/JobManager/add. If you want to update an already transferred job, for example to change the state, simply transfer the job again. PERSEUS will auto-detect if the job is already in the database by checking the job_id as well as the tuple group_id and group_index. If a match is found, it will overwrite the existing job instead of creating a new one.

Attributes

A job contains the following information:

Attribute	Type	Description
job_id*	int	The job id given by the scheduler
group_id	int	In case this job is a group / array job, the group’s id
group_index	int	In case this job is a group / array job, the corresponding index
job_name	str	A human readable job name
project_oid*	str	The ObjectId of the corresponding PERSEUS project
compute_project_id*	str	The identifier of the corresponding compute project
cluster_id*	str	The identifier of the cluster the job was submitted to (has to be identical with the Cluster.id property within PERSEUS)
partition	str	The partition the job was submitted to
nodes	list[str]	The partition the nodes was submitted to
allocated_resources	list[str]	The allocated resources of the job as human readable strings
working_directory	str	The working directory of the job
submitted	str	An ISO-formatted datetime string indicating when the job was submitted
start	str	An ISO-formatted datetime string indicating when the job started (or will start)
end	str	An ISO-formatted datetime string indicating when the job ended (or will end)
time_limit	int	Given time limit in seconds
user	str	The username of the user who submitted the job (to benefit from all features, the username should be a property known to PERSEUS by the Person.username property)
state	str	The current state in which the job is in (see job states for further information)
priority_id	str	The identifier of the priority used for submitting the job (see how to manage priorities for further information)
reservation	str	The reservation used for job submission
reason	str	A reason for non-successful job submission / finish
cluster_cockpit_id	int	The identifier from cluster cockpit referencing this job
additional_data	dict[str, Any]	Any data you want to save together with this job, for example further statistics

Please see Services - JobManager in the API documentation for details on the related data models.

Job states

The following states are available as job state. See job states in Andromeda for further reference.

State	Meaning
RUNNING	The job has resources allocated and is currently running.
PENDING	The job is queued and waiting for resources to become available.
COMPLETED	The job finished successfully.
SUSPENDED	The job has an allocation but its execution is currently paused.
CANCELLED	The job was cancelled by a user or an administrator.
BOOT_FAIL	The allocated nodes failed to boot, so the job could not start.
DEADLINE	The job reached its configured deadline before it could complete.
FAILED	The job terminated with a non-zero exit code.
NODE_FAIL	One or more allocated nodes failed while the job was running.
OUT_OF_MEMORY	The job exceeded the memory available to it.
PREEMPTED	The job was stopped so higher-priority work could use its resources.
TIMEOUT	The job ran longer than its time limit and was stopped.