'Invalid Trackable RESource (TRES) specification' error

[kimchitsigai]

Hi,

I'm using slurm-drmaa to submit a job and I get the error below:

d #89f27 [     0.00]  * # Native specification:  --time=1:00:00 --ntasks=1 --gres=gpu:1 --cpus-per-task=2 --nodes=1 --account=xxx@yyy --partition=mypartition

t #89f27 [     0.00] -> slurmdrmaa_parse_native

t #89f27 [     0.00] -> slurmdrmaa_parse_additional_attr

d #89f27 [     0.00]  * # time_limit = 1:00:00

t #89f27 [     0.00] -> slurmdrmaa_datetime_parse(1:00:00)

d #89f27 [     0.00]  * parsed: 0000-00-00 01:00:00 +00:00:00 [---hms-]

t #89f27 [     0.00] <- slurmdrmaa_datetime_parse(1:00:00) =60 minutes
t #89f27 [     0.00] <- slurmdrmaa_parse_additional_attr

t #89f27 [     0.00] -> slurmdrmaa_parse_additional_attr
d #89f27 [     0.00]  * # ntasks = 1
t #89f27 [     0.00] <- slurmdrmaa_parse_additional_attr

t #89f27 [     0.00] -> slurmdrmaa_parse_additional_attr
d #89f27 [     0.00]  * # gres = gpu:1
t #89f27 [     0.00] <- slurmdrmaa_parse_additional_attr

t #89f27 [     0.00] -> slurmdrmaa_parse_additional_attr
d #89f27 [     0.00]  * # cpus_per_task = 2
t #89f27 [     0.00] <- slurmdrmaa_parse_additional_attr

t #89f27 [     0.00] -> slurmdrmaa_parse_additional_attr
d #89f27 [     0.00]  * nodes: 1 ->
d #89f27 [     0.00]  * # min_nodes = 1
t #89f27 [     0.00] <- slurmdrmaa_parse_additional_attr

t #89f27 [     0.00] -> slurmdrmaa_parse_additional_attr
d #89f27 [     0.00]  * # account = xxx@yyy
t #89f27 [     0.00] <- slurmdrmaa_parse_additional_attr

t #89f27 [     0.00] -> slurmdrmaa_parse_additional_attr
d #89f27 [     0.00]  * # partition = allgpus
t #89f27 [     0.00] <- slurmdrmaa_parse_additional_attr

d #89f27 [     0.00]  * finalizing job constraints
d #89f27 [     0.00]  * set min_cpus to ntasks*cpus_per_task: 2
t #89f27 [     0.00] <- slurmdrmaa_parse_native
E #89f27 [     4.24]  * fsd_exc_new(1016,slurm_submit_batch_job error: Invalid Trackable RESource (TRES) specification,1)

t #89f27 [     4.24] -> slurmdrmaa_free_job_desc
t #89f27 [     4.24] <- slurmdrmaa_free_job_desc

t #89f27 [     4.24] <- drmaa_run_job=17: slurm_submit_batch_job error: Invalid Trackable RESource (TRES) specification

Traceback (most recent call last):
  ..
  File "/.../python3.6/site-packages/drmaa/session.py", line 314, in runJob
    c(drmaa_run_job, jid, sizeof(jid), jobTemplate)
  File "/.../python3.6/site-packages/drmaa/helpers.py", line 302, in c
    return f(*(args + (error_buffer, sizeof(error_buffer))))
  File "/.../site-packages/drmaa/errors.py", line 151, in error_check
    raise _ERRORS[code - 1](error_string)
drmaa.errors.DeniedByDrmException: code 17: slurm_submit_batch_job error: Invalid Trackable RESource (TRES) specification

The same job without `--gres=gpu:1 works fine. slurm-drmaa version is 1.1.3 and slurm version is 21.08.6. Os is RHEL 8.4.

Any hint would be greatly appreciated, Kimchi

[kimchitsigai]

Hi,

I've added a logging instruction in slurm_drmaa/session.c: fsd_log_debug(("job_desc.tres_per_node = %s", job_desc.tres_per_node)); and it produced this output: job_desc.tres_per_node = gpu:1 And finally, I get the Invalid Trackable (TRES) specification RESource error.

By changing the input from --gres=gpu:1 to --gres=gres:gpu:1, the logging instruction outputs: job_desc.tres_per_node = gres:gpu:1 as expected. The TRES error disappeared and the job was correctly submitted.

By looking at the code in https://github.com/SchedMD/slurm/blob/master/src/common/gres.c and https://github.com/SchedMD/slurm/blob/master/src/common/slurm_opt.c it seems that Slurm is expecting GPU resource requests formatted as gres:gpu:1 and not as gpu:1 as in the previous versions (20.x) of Slurm.

Shouldn't slurm-drmaa be updated to take into account this change in Slurm?

Best, Kimchi

[scholtalbers]

I can confirm that changing --gres=gpu:1 to --gres=gres:gpu:1 solves the TRES error.

[natefoo]

Thanks, yes, it sounds like we should probably prepend gres: to the string in the job template. gres in the job template was generalized to tres_per_node in 18.08, I am not sure if this is when it broke.