oh-my-batch 0.1.0.dev0__py3-none-any.whl → 0.1.0.dev2__py3-none-any.whl

oh_my_batch/job.py

@@ -12,29 +12,22 @@ from .util import expand_globs, shell_run, parse_csv
 
 logger = logging.getLogger(__name__)
 
+class JobState:
+    NULL = 0
+    PENDING = 1
+    RUNNING = 2
+    CANCELLED = 3
+    COMPLETED = 4
+    FAILED = 5
+    UNKNOWN = 6
 
-class JobState(bytes, Enum):
-    """
-    Job state enumeration
-    """
-    def __new__(cls, value: int, terminal: bool, status_name: str) -> "JobState":
-        obj = bytes.__new__(cls, [value])
-        obj._value_ = value
-        obj.terminal = terminal
-        obj.status_name = status_name
-        return obj
-
-    value: int  # type: ignore
-    terminal: bool
-    status_name: str
-
-    NULL = (0, True, "NULL")
-    PENDING = (1, False, "PENDING")
-    RUNNING = (2, False, "RUNNING")
-    CANCELLED = (3, True, "CANCELLED")
-    COMPLETED = (4, True, "COMPLETED")
-    FAILED = (5, True, "FAILED")
-    UNKNOWN = (6, False, "UNKNOWN")
+    @classmethod
+    def is_terminal(cls, state: int):
+        return state in (JobState.NULL, JobState.COMPLETED, JobState.FAILED, JobState.CANCELLED)
+
+    @classmethod
+    def is_success(cls, state: int):
+        return state == JobState.COMPLETED
 
 
 def new_job(script: str):
@@ -69,9 +62,6 @@ class BaseJobManager:
         scripts = set(os.path.normpath(s) for s in expand_globs(script))
         logger.info('Scripts to submit: %s', scripts)
 
-        if recovery and recover_scripts != scripts:
-            raise ValueError('Scripts to submit are different from scripts in recovery file')
-
         for script_file in scripts:
             if script_file not in recover_scripts:
                 jobs.append(new_job(script_file))
@@ -87,7 +77,7 @@ class BaseJobManager:
                 break
 
             # stop if all jobs are terminal and not job to be submitted
-            if (all(j['state'].terminal for j in jobs) and
+            if (all(JobState.is_terminal(j['state']) for j in jobs) and
                     not any(should_submit(j, max_tries) for j in jobs)):
                 break
 
@@ -108,19 +98,21 @@ class Slurm(BaseJobManager):
 
     def _update_jobs(self, jobs: List[dict], max_tries: int, submit_opts: str):
         # query job status
-        job_ids = ','.join(j['id'] for j in jobs if j['id'])
-        query_cmd = f'{self._sacct_bin} -X -P -j {job_ids} --format=JobID,JobName,State'
-
-        user = os.environ.get('USER')
-        if user:
-            query_cmd += f' -u {user}'
-
-        cp = shell_run(query_cmd)
-        if cp.returncode != 0:
-            logger.error('Failed to query job status: %s', cp.stderr.decode('utf-8'))
-            return jobs
-        logger.info('Job status: %s', cp.stdout.decode('utf-8'))
-        new_state = parse_csv(cp.stdout.decode('utf-8'))
+        job_ids = [j['id'] for j in jobs if j['id']]
+        if job_ids:
+            query_cmd = f'{self._sacct_bin} -X -P --format=JobID,JobName,State -j {",".join(job_ids)}'
+            user = os.environ.get('USER')
+            if user:
+                query_cmd += f' -u {user}'
+
+            cp = shell_run(query_cmd)
+            if cp.returncode != 0:
+                logger.error('Failed to query job status: %s', cp.stderr.decode('utf-8'))
+                return jobs
+            logger.info('Job status: %s', cp.stdout.decode('utf-8'))
+            new_state = parse_csv(cp.stdout.decode('utf-8'))
+        else:
+            new_state = []
 
         for job in jobs:
             for row in new_state:
@@ -130,8 +122,8 @@ class Slurm(BaseJobManager):
                         logger.warning('Unknown job %s state: %s',row['JobID'], row['State'])
                     break
             else:
-                job['state'] = JobState.FAILED
-                logger.error('Job %s not found in sacct output', job['id'])
+                if job['id']:
+                    logger.error('Job %s not found in sacct output', job['id'])
 
         # check if there are jobs to be (re)submitted
         for job in jobs:
@@ -171,8 +163,8 @@ class Slurm(BaseJobManager):
 
 
 def should_submit(job: dict, max_tries: int):
-    state: JobState = job['state']
-    if not state.terminal:
+    state: int = job['state']
+    if not JobState.is_terminal(state):
         return False
     if job['tries'] >= max_tries:
         return False

{oh_my_batch-0.1.0.dev0.dist-info → oh_my_batch-0.1.0.dev2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: oh-my-batch
-Version: 0.1.0.dev0
+Version: 0.1.0.dev2
 Summary: 
 License: GPL
 Author: weihong.xu
@@ -118,13 +118,13 @@ You can use `omb job` to track the state of the jobs.
 ```bash
 
 omb job slurm \
-    submit tmp/*.slurm --max_tries 3 --wait --recovery lammps.recovery
+    submit tmp/*.slurm --max_tries 3 --wait --recovery lammps-jobs.json 
 ```
 
 The above command will submit the batch scripts to the job scheduler,
 and wait for the jobs to finish. If the job fails, it will retry for at most 3 times.
 
-The `--recovery` option will save the job information to `lammps.recovery` file,
+The `--recovery` option will save the job information to `lammps-jobs.json` file,
 if `omb job` is interrupted, you can run the exact same command to recover the job status, 
 so that you don't need to resubmit the jobs that are already submitted.
 

{oh_my_batch-0.1.0.dev0.dist-info → oh_my_batch-0.1.0.dev2.dist-info}/RECORD

@@ -5,10 +5,10 @@ oh_my_batch/assets/functions.sh,sha256=eORxFefV-XrWbG-2I6u-c8uf1XxOQ31LaeVHBumwz
 oh_my_batch/batch.py,sha256=e73N-xwxMvgxnWwFMp33PQD1Dy-T-ATjANlwtPRHPQM,3016
 oh_my_batch/cli.py,sha256=G_JxqX0Zbx_EbcDxXbYjJ_4O-EOhmkF1lcMWgQ5ZPqo,375
 oh_my_batch/combo.py,sha256=AHFD5CLoczqtjcfl2Rb4A2ucoQU40-cWtDOYjtP-yY4,7680
-oh_my_batch/job.py,sha256=kup6Kwr3HFeCWAYJzJ1BET81_Dvbz1HxuHfmMPOpCnU,6080
+oh_my_batch/job.py,sha256=_fETBYpuSd_hNHKnXSwYcSU3OXtU7PO-P2QMfhE-Wfs,5788
 oh_my_batch/util.py,sha256=H8B4zVNH5xRp-NG_uypgvtmz2YSpXy_6LK5ROv6SYrc,2116
-oh_my_batch-0.1.0.dev0.dist-info/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
-oh_my_batch-0.1.0.dev0.dist-info/METADATA,sha256=nUAgfApBgfXFhHD9-VWXAZsyGF1iJcO8bxYGMZKcGLI,4453
-oh_my_batch-0.1.0.dev0.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-oh_my_batch-0.1.0.dev0.dist-info/entry_points.txt,sha256=ZY2GutSoNjjSyJ4qO2pTeseKUFgoTYdvmgkuZZkwi68,77
-oh_my_batch-0.1.0.dev0.dist-info/RECORD,,
+oh_my_batch-0.1.0.dev2.dist-info/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
+oh_my_batch-0.1.0.dev2.dist-info/METADATA,sha256=Novvp5-MhKR9J0Q6Al833vA6GjT7EVRtDv4ADmnluxk,4456
+oh_my_batch-0.1.0.dev2.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+oh_my_batch-0.1.0.dev2.dist-info/entry_points.txt,sha256=ZY2GutSoNjjSyJ4qO2pTeseKUFgoTYdvmgkuZZkwi68,77
+oh_my_batch-0.1.0.dev2.dist-info/RECORD,,