Merge pull request #369 from bedroge/fix_scontrol_output_parsing

Improved parsing of `scontrol` output

Author: trz42

eessi_bot_job_manager.py

@@ -272,6 +272,9 @@ def parse_scontrol_show_job_output(self, output):
         """
         The output of 'scontrol --oneliner show job' is a list of key=value pairs
         separated by whitespaces.
+        Note that with newer Slurm versions (25.11), some values can also contain whitespaces
+        (e.g. for SubmitLine), making it complex to distinguish between keys and values.
+        To solve this, we assume that all Slurm keys start with an uppercase letter.
 
         Args:
             output (string): the output of the scontrol command
@@ -281,8 +284,20 @@ def parse_scontrol_show_job_output(self, output):
         """
         job_info = {}
         stripped_output = output.strip()
-        for pair in stripped_output.split():
-            key, value = pair.split('=', 1)
+
+        # Match keys that start with uppercase and continue until '='
+        key_pattern = re.compile(r'([A-Z][A-Za-z0-9:/_]*)=')
+
+        keys_matches = list(key_pattern.finditer(stripped_output))
+
+        for idx, key_match in enumerate(keys_matches):
+            # The key is what actually got matched by the regex
+            key = key_match.group(1)
+            # The value starts where the key ends...
+            value_start = key_match.end()
+            # and it ends where the next key starts (or, if it's the last one, where the string ends)
+            value_end = keys_matches[idx+1].start() if (idx + 1) < len(keys_matches) else len(stripped_output)
+            value = stripped_output[value_start:value_end].strip()
             job_info[key] = value
 
         return job_info

tests/test_eessi_bot_job_manager.py

@@ -135,3 +135,31 @@ def test_determine_finished_jobs():
     assert job_manager.determine_finished_jobs(known_jobs, current_jobs_all_jobs) == []
     assert job_manager.determine_finished_jobs(known_jobs, current_jobs_one_job) == ['1', '2']
     assert job_manager.determine_finished_jobs(known_jobs, {}) == ['0', '1', '2']
+
+
+def test_parse_scontrol_show_job_output():
+    # Dummy output (shortened) from Slurm 25.11.3 for "scontrol show job <jobid>"
+    scontrol_output = 'JobId=123 JobName=bot_test_job UserId=eessibot(12345) MCS_label=N/A EligibleTime=Unknown' \
+                      ' AllocNode:Sid=my.node.name:123456 SubmitLine=/opt/slurm/25.11.3/bin/sbatch --hold' \
+                      ' --time=10-0:0:0 --nodes=1 --exclusive --cpus-per-task=1 --job-name=bot_test_job ' \
+                      '/home/eessibot/job.slurm WorkDir=/jobs/2026.01/pr_123/event_123-456-789/run_000/riscv64/' \
+                      'generic/dev.eessi.io-riscv StdErr= StdIn=/dev/null StdOut=/jobs/2026.01/pr_123/' \
+                      'event_123-456-789/run_000/riscv64/generic/dev.eessi.io-riscv/slurm-123.out TresPerTask=cpu=1'
+    job_manager = EESSIBotSoftwareLayerJobManager()
+    job_info = job_manager.parse_scontrol_show_job_output(scontrol_output)
+    job_info_expected = {
+        'JobId': '123',
+        'JobName': 'bot_test_job',
+        'UserId': 'eessibot(12345)',
+        'MCS_label': 'N/A',
+        'EligibleTime': 'Unknown',
+        'AllocNode:Sid': 'my.node.name:123456',
+        'SubmitLine': '/opt/slurm/25.11.3/bin/sbatch --hold --time=10-0:0:0 --nodes=1 --exclusive --cpus-per-task=1 '
+                      '--job-name=bot_test_job /home/eessibot/job.slurm',
+        'WorkDir': '/jobs/2026.01/pr_123/event_123-456-789/run_000/riscv64/generic/dev.eessi.io-riscv',
+        'StdErr': '',
+        'StdIn': '/dev/null',
+        'StdOut': '/jobs/2026.01/pr_123/event_123-456-789/run_000/riscv64/generic/dev.eessi.io-riscv/slurm-123.out',
+        'TresPerTask': 'cpu=1',
+    }
+    assert job_info == job_info_expected