Merge pull request #1746 from buildtesters/change_pbs_and_torque_poll…

…_commands_and_jobdata

Change Poll commands for PBS and Torque and add support for retrieving job data

buildtest/buildsystem/builders.py

@@ -400,7 +400,6 @@ def _skip_tests_by_tags(self, recipe, name):
  if not recipe.get("tags"):
  return True
 
- found = False
  # the input tag names from test can be list or string
  tests_in_tags = recipe["tags"]
  # if input is string, convert to list otherwise we assume its a list

buildtest/scheduler/pbs.py

@@ -1,3 +1,4 @@
+import json
 import logging
 import os
 import re
@@ -15,6 +16,8 @@ class PBSJob(Job):
  to poll job state, gather job results upon completion and cancel job.
  """
 
+ poll_command = "qstat -xf"
+
  def __init__(self, jobID):
  self._outfile = None
  self._errfile = None
@@ -54,7 +57,7 @@ def fail(self):
 
  def get_output_error_files(self):
  """Fetch output and error files right after job submission."""
- query = f"qstat -f {self.jobid}"
+ query = f"{self.poll_command} {self.jobid}"
  cmd = BuildTestCommand(query)
  cmd.execute()
  output = " ".join(cmd.get_output())
@@ -198,7 +201,7 @@ def poll(self):
  copy_on_rerun = False
 
  """
- query = f"qstat -f {self.jobid}"
+ query = f"{self.poll_command} {self.jobid}"
 
  logger.debug(f"Polling job by running: {query}")
  cmd = BuildTestCommand(query)
@@ -252,14 +255,16 @@ def retrieve_jobdata(self):
  for getting output file, error file and exit status of job.
  """
 
- # job_data = json.loads(output)
+ query = f"{self.poll_command} -F json {self.jobid}"
+ logger.debug(f"Retrieving job data by running: {query}")
+ cmd = BuildTestCommand(query)
+ cmd.execute()
+ output = " ".join(cmd.get_output())
+ job_data = json.loads(output)
 
- # output in the form of pbs:<path>
- # self._outfile = job_data["Jobs"][self.jobid]["Output_Path"].split(":")[1]
- # self._errfile = job_data["Jobs"][self.jobid]["Error_Path"].split(":")[1]
+ logger.debug(f"Retrieved job data for job: {self.jobid}:\n{job_data}")
 
- # self._exitcode = self._get_exitcode()
- return {}
+ return job_data
 
  def cancel(self):
  """Cancel PBS job by running ``qdel <jobid>``."""
@@ -270,4 +275,19 @@ def cancel(self):
 
 
 class TorqueJob(PBSJob):
- pass
+
+ poll_command = "qstat -f"
+
+ def retrieve_jobdata(self):
+ """This method is called once job is complete. We will gather record of job by running
+ `qstat -f <jobid>` and return the output as a string.
+ """
+ query = f"{self.poll_command} {self.jobid}"
+ logger.debug(f"Retrieving job data by running: {query}")
+ cmd = BuildTestCommand(query)
+ cmd.execute()
+ output = " ".join(cmd.get_output())
+
+ logger.debug(f"Retrieved job data for job: {self.jobid}:\n{output}")
+
+ return output

tests/test_pbs.py

@@ -1,18 +1,19 @@
 import os
-import shutil
 
 import pytest
 
 from buildtest.cli.build import BuildTest
 from buildtest.config import SiteConfiguration
+from buildtest.scheduler.detection import PBS
 from buildtest.system import BuildTestSystem
 from buildtest.utils.file import walk_tree
 
 
 def test_pbs():
  """Need to figure out a PBS environment where to run this regression test."""
- if not shutil.which("pbsnodes"):
- pytest.skip("Test runs only on PBS Cluster, must have pbsnodes command")
+ pbs = PBS()
+ if not pbs.active():
+ pytest.skip("Test runs only on PBS Cluster")
 
  here = os.path.dirname(os.path.abspath(__file__))
  settings_file = os.path.join(here, "settings", "pbs.yml")