static int _job_signal(uint32_t jobid, uint16_t sig_num)
{
struct job_record *job_ptr;
int rc = SLURM_SUCCESS;
job_ptr = find_job_record(jobid);
if (job_ptr == NULL)
return ESLURM_INVALID_JOB_ID;
if (IS_JOB_FINISHED(job_ptr))
return ESLURM_ALREADY_DONE;
if (job_ptr->batch_flag)
rc = job_signal(jobid, sig_num, 1, 0, false);
if (rc == SLURM_SUCCESS)
rc = job_signal(jobid, sig_num, 0, 0, false);
return rc;
}
static void _preempt_job_dequeue(void)
{
struct job_record *job_ptr;
uint32_t job_id, *tmp_id;
uint16_t preempt_mode;
xassert(preempt_job_list);
while ((tmp_id = list_pop(preempt_job_list))) {
int rc = SLURM_ERROR;
job_id = *tmp_id;
xfree(tmp_id);
if ((job_ptr = find_job_record(job_id)) == NULL) {
error("_preempt_job_dequeue could not find job %u",
job_id);
continue;
}
preempt_mode = slurm_job_preempt_mode(job_ptr);
if (preempt_mode == PREEMPT_MODE_SUSPEND) {
if ((rc = _suspend_job(job_id)) == ESLURM_DISABLED)
rc = SLURM_SUCCESS;
} else if (preempt_mode == PREEMPT_MODE_CANCEL) {
rc = job_signal(job_ptr->job_id, SIGKILL, 0, 0, true);
if (rc == SLURM_SUCCESS) {
info("preempted job %u has been killed",
job_ptr->job_id);
}
} else if (preempt_mode == PREEMPT_MODE_CHECKPOINT) {
checkpoint_msg_t ckpt_msg;
memset(&ckpt_msg, 0, sizeof(checkpoint_msg_t));
ckpt_msg.op = CHECK_REQUEUE;
ckpt_msg.job_id = job_ptr->job_id;
rc = job_checkpoint(&ckpt_msg, 0, -1,
(uint16_t)NO_VAL);
if (rc == ESLURM_NOT_SUPPORTED) {
memset(&ckpt_msg, 0, sizeof(checkpoint_msg_t));
ckpt_msg.op = CHECK_VACATE;
ckpt_msg.job_id = job_ptr->job_id;
rc = job_checkpoint(&ckpt_msg, 0, -1,
(uint16_t)NO_VAL);
}
if (rc == SLURM_SUCCESS) {
info("preempted job %u has been checkpointed",
job_ptr->job_id);
} else
error("preempted job %u could not be "
"checkpointed: %s",
job_ptr->job_id, slurm_strerror(rc));
} else if ((preempt_mode == PREEMPT_MODE_REQUEUE) &&
job_ptr->batch_flag && job_ptr->details &&
(job_ptr->details->requeue > 0)) {
rc = job_requeue(0, job_ptr->job_id, -1,
(uint16_t)NO_VAL, true, 0);
if (rc == SLURM_SUCCESS) {
info("preempted job %u has been requeued",
job_ptr->job_id);
} else
error("preempted job %u could not be "
"requeued: %s",
job_ptr->job_id, slurm_strerror(rc));
}
if (rc != SLURM_SUCCESS) {
rc = job_signal(job_ptr->job_id, SIGKILL, 0, 0, true);
if (rc == SLURM_SUCCESS)
info("preempted job %u had to be killed",
job_ptr->job_id);
else {
info("preempted job %u kill failure %s",
job_ptr->job_id, slurm_strerror(rc));
}
}
}
return;
}