Added support for STAGE_OUT mail type for issue #36

neilmunday · May 9, 2022 · 7797931 · 7797931
1 parent e4b4141
commit 7797931
Show file tree

Hide file tree

Showing 5 changed files with 42 additions and 9 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -16,7 +16,7 @@ Date: 2022-05-09
 * Changed spool mail file to use JSON format.
 * Moved templates into conf.d/templates.
 * Created e-mail signature template.
-* Added support for REQUEUE, INVALID_DEPEND, TIME_LIMIT_50, TIME_LIMIT_80, TIME_LIMIT_90, and TIME_LIMIT mail types (issue #36).
+* Added support for REQUEUE, INVALID_DEPEND, STAGE_OUT, TIME_LIMIT_50, TIME_LIMIT_80, TIME_LIMIT_90, and TIME_LIMIT mail types (issue #36).
 
 Version 2.5
 -----------

diff --git a/README.md b/README.md
@@ -117,6 +117,7 @@ Slurm-Mail uses Python's [string.Template](https://docs.python.org/3/library/str
 | invalid-dependency        | Used when a job has an invalid dependency.                        |
 | job_table.tpl             | Used to create the job info table in e-mails.                     |
 | signature.tpl             | Used to create the e-mail signature.                              |
+| staged-out.tpl            | Used when a job's burst buffer stage has completed.               |
 | started.tpl               | Used for jobs that have started.                                  |
 | started-array-summary.tpl | Used when the first job in an array has started.                  |
 | started-array.tpl         | Used for the first job in an array that has started.              |

diff --git a/bin/slurm-send-mail.py b/bin/slurm-send-mail.py
@@ -362,6 +362,7 @@ def process_spool_file(json_file: pathlib.Path):
             "Failed",
             "Invalid dependency",
             "Requeued",
+            "Staged Out",
             "Time reached 50%",
             "Time reached 80%",
             "Time reached 90%",
@@ -529,12 +530,6 @@ def process_spool_file(json_file: pathlib.Path):
                     USER=pwd.getpwnam(job.user).pw_gecos, JOB_TABLE=job_table,
                     CLUSTER=job.cluster
                 )
-        elif state == "Invalid dependency":
-            tpl = Template(get_file_contents(templates['invalid_dependency']))
-            body = tpl.substitute(
-                CSS=css, CLUSTER=job.cluster, JOB_ID=job.id, SIGNATURE=signature,
-                USER=pwd.getpwnam(job.user).pw_gecos, JOB_TABLE=job_table,
-            )
         elif state in ["Ended", "Failed", "Requeued", "Time limit reached"]:
             end_txt = state.lower()
             job_output = ""
@@ -596,6 +591,18 @@ def process_spool_file(json_file: pathlib.Path):
             )
             # change state value for upcomming e-mail send
             state = "{0}% of time limit reached".format(reached)
+        elif state == "Invalid dependency":
+            tpl = Template(get_file_contents(templates['invalid_dependency']))
+            body = tpl.substitute(
+                CSS=css, CLUSTER=job.cluster, JOB_ID=job.id, SIGNATURE=signature,
+                USER=pwd.getpwnam(job.user).pw_gecos, JOB_TABLE=job_table,
+            )
+        elif state == "Staged Out":
+            tpl = Template(get_file_contents(templates['staged_out']))
+            body = tpl.substitute(
+                CSS=css, CLUSTER=job.cluster, JOB_ID=job.id, SIGNATURE=signature,
+                USER=pwd.getpwnam(job.user).pw_gecos, JOB_TABLE=job_table,
+            )
 
         msg = MIMEMultipart("alternative")
         msg['subject'] = Template(email_subject).substitute(
@@ -695,6 +702,7 @@ def tail_file(f: str, num_lines: int) -> str:
     templates['job_output'] = tpl_dir / "job-output.tpl"
     templates['job_table'] = tpl_dir / "job-table.tpl"
     templates['signature'] = tpl_dir / "signature.tpl"
+    templates['staged_out'] = tpl_dir / "staged-out.tpl"
     templates['started'] = tpl_dir / "started.tpl"
     templates['time'] = tpl_dir / "time.tpl"
 

diff --git a/bin/slurm-spool-mail.py b/bin/slurm-spool-mail.py
@@ -120,15 +120,15 @@ def die(msg: str):
         match = None
         if "Array" in info:
             match = re.search(
-                r"Slurm ((?P<array_summary>Array Summary)|Array Task) Job_id=[0-9]+_([0-9]+|\*) \((?P<job_id>[0-9]+)\).*?(?P<state>(Began|Ended|Failed|Requeued|Invalid dependency|Reached time limit|Reached (?P<limit>[0-9]+)% of time limit))",
+                r"Slurm ((?P<array_summary>Array Summary)|Array Task) Job_id=[0-9]+_([0-9]+|\*) \((?P<job_id>[0-9]+)\).*?(?P<state>(Began|Ended|Failed|Requeued|Invalid dependency|Reached time limit|Reached (?P<limit>[0-9]+)% of time limit|Staged Out))",
                 info
             )
             if not match:
                 die("Failed to parse Slurm info.")
             array_summary = (match.group("array_summary") is not None)
         else:
             match = re.search(
-                r"Slurm Job_id=(?P<job_id>[0-9]+).*?(?P<state>(Began|Ended|Failed|Requeued|Invalid dependency|Reached time limit|Reached (?P<limit>[0-9]+)% of time limit))",
+                r"Slurm Job_id=(?P<job_id>[0-9]+).*?(?P<state>(Began|Ended|Failed|Requeued|Invalid dependency|Reached time limit|Reached (?P<limit>[0-9]+)% of time limit|Staged Out))",
                 info
             )
             if not match:

diff --git a/conf.d/templates/staged-out.tpl b/conf.d/templates/staged-out.tpl
@@ -0,0 +1,24 @@
+<html>
+<head>
+<style>
+	$CSS
+
+	tr.jobEnd {
+		display: none;
+	}
+</style>
+</head>
+<body>
+
+<p>Dear $USER,</p>
+
+<p>Your job $JOB_ID on $CLUSTER has completed its burst buffer stage out.</p>
+
+<p>Details about the job can be found in the table below:</p>
+
+$JOB_TABLE
+
+$SIGNATURE
+
+</body>
+</html>