Initial commit

pyproject.toml

@@ -0,0 +1,13 @@
+[build-system]
+requires = ["setuptools", "systemd-python"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "sysalert"
+version = "0.1.0"
+dependencies = [
+  "systemd-python"
+]
+
+[project.scripts]
+sysalert = "sysalert.util:cli"

sysalert/db.py

@@ -0,0 +1,129 @@
+import datetime
+import os
+import sqlite3
+
+def register_success(name, db):
+    cur = db.cursor()
+    cur.execute('''
+            DELETE FROM alert
+                WHERE failure IN (
+                    SELECT c.id
+                        FROM failure c INNER JOIN service s
+                            ON c.service=s.id
+                        WHERE s.name=?
+                        );
+            ''', (name,))
+    cur.execute('''
+            DELETE FROM failure
+                WHERE service IN (
+                    SELECT id FROM service WHERE name=?
+                    );
+                ''', (name,))
+    cur.execute('''
+            DELETE FROM service WHERE name=?;
+            ''', (name,))
+    db.commit()
+
+def register_failure(name, alert_method, db):
+    cur = db.cursor()
+    cur.execute('INSERT OR IGNORE INTO service(name) VALUES (?);', (name,))
+    cur.execute('SELECT id FROM service WHERE name=?;', (name,))
+    (service_id,) = cur.fetchone()
+    cur.execute(''' INSERT INTO failure
+            ( service, service_result, exit_code, exit_status, invocation_id, timestamp)
+            VALUES
+            ( ?,       ?,              ?,         ?,           ?,             ?)
+            RETURNING id;
+            ''', (
+                service_id,
+                os.environ['MONITOR_SERVICE_RESULT'],
+                os.environ['MONITOR_EXIT_CODE'],
+                os.environ['MONITOR_EXIT_STATUS'],
+                os.environ['MONITOR_INVOCATION_ID'],
+                datetime.datetime.now()
+                ))
+    if alert_method:
+        (failure_id,) = cur.fetchone()
+        cur.execute(''' INSERT INTO alert
+                (failure, timestamp, method)
+                VALUES
+                (?,       ?,         ?);
+                ''', (
+                    failure_id,
+                    datetime.datetime.now(),
+                    alert_method
+                    ))
+    db.commit()
+
+
+
+def get_failures(name, db):
+    cur = db.cursor()
+    cur.execute('''SELECT
+            failure.id, service_result, exit_code, exit_status, invocation_id, timestamp
+            FROM failure
+                INNER JOIN service ON
+                    failure.service=service.id
+            WHERE service.name=?;
+                ''', (name,))
+
+    failures=[]
+    for f in cur.fetchall():
+        cur.execute('SELECT method FROM alert where failure=?', (f[0],))
+        alert_method = cur.fetchone()
+        if alert_method:
+            alert_method = alert_method[0]
+        failures.append({
+            'service_result': f[1],
+            'exit_code': f[2],
+            'exit_status': f[3],
+            'invocation_id': f[4],
+            'timestamp': f[5],
+            'alert_method': alert_method
+            })
+
+    failures.sort(key=lambda x: x['timestamp'])
+    return failures
+
+
+def init(path):
+    con = sqlite3.connect(path, detect_types=sqlite3.PARSE_DECLTYPES | sqlite3.PARSE_COLNAMES)
+    cur = con.cursor()
+    cur.execute('''
+            CREATE TABLE IF NOT EXISTS
+                service (
+                    id INTEGER PRIMARY KEY AUTOINCREMENT,
+                    name TEXT UNIQUE
+                    );
+                ''')
+    cur.execute('''
+            CREATE TABLE IF NOT EXISTS
+                failure (
+                    id INTEGER PRIMARY KEY AUTOINCREMENT,
+                    service INTEGER,
+                    service_result TEXT,
+                    exit_code TEXT,
+                    exit_status INTEGER,
+                    invocation_id TEXT,
+                    timestamp timestamp,
+
+                    FOREIGN KEY(service) REFERENCES service(id)
+                    );
+                ''')
+    cur.execute('''
+            CREATE TABLE IF NOT EXISTS
+                alert (
+                    id INTEGER PRIMARY KEY AUTOINCREMENT,
+                    failure INTEGER,
+                    timestamp timestamp,
+                    method TEXT,
+
+                    FOREIGN KEY(failure) REFERENCES failure(id)
+                    );
+                ''')
+    con.commit();
+    return con
+
+def close(con):
+    con.close()
+

sysalert/email.py

@@ -0,0 +1,98 @@
+import email
+import smtplib
+import socket
+import subprocess
+
+import systemd.journal
+
+def _get_status(name):
+    cmd = ['/usr/bin/systemctl', 'status', name]
+    
+    proc = subprocess.Popen(
+	    cmd,
+	    stdout=subprocess.PIPE)
+    out, err = proc.communicate()
+    out = out.decode('utf-8')
+    return out
+
+def _send_email(subject, message, config):
+    mail = email.message.EmailMessage()
+    hostname = socket.getfqdn()
+    smtp_host='localhost'
+    mail_from=f'sysalert <root@{hostname}>'
+    mail_to=f'root <root@{hostname}>'
+    if 'smtp_host' in config:
+        smtp_host = config['smtp_host']
+    if 'mail_from' in config:
+        mail_from = config['mail_from']
+    if 'mail_to' in config:
+        mail_to = config['mail_to']
+
+
+    mail.set_content(message)
+    mail['To'] = mail_to
+    mail['From'] = mail_from
+    mail['Subject'] = subject
+
+    with smtplib.SMTP(smtp_host) as smtp:
+        smtp.send_message(mail)
+
+def success(name, failures, config):
+    cur_status = _get_status(name)
+    nr_failures = len(failures)
+    hostname = socket.getfqdn()
+    subject=f'{hostname} - {name}: recovery'
+    message=f"""
+{name} has recovered after {nr_failures} failures.
+Current status:
+
+{cur_status}
+"""
+    _send_email(subject, message, config)
+
+
+def failure(name, failures, config):
+    cur_status = _get_status(name)
+    hostname = socket.getfqdn()
+    latest_inv_id = failures[-1]['invocation_id']
+    nr_failures = len(failures)
+    prev_failures=0
+    for i,f in enumerate(reversed(failures[:-1])):
+        if f['alert_method']:
+            prev_failures=i+1
+            break
+
+    reader = systemd.journal.Reader()
+    reader.add_match(MONITOR_INVOCATION_ID=latest_inv_id)
+    reader.add_disjunction()
+    reader.add_match(_SYSTEMD_INVOCATION_ID=lateset_inv_id)
+    journal_txt = "\n".join([entry['MESSAGE'] for entry in reader])
+
+    if nr_failures <= 1:
+        subject=f"{hostname} - {name}: failure"
+        message=f"""
+New failure for {name}.
+Current status:
+
+{cur_status}
+
+
+Latest journal log:
+
+{journal_txt}
+"""
+    else:
+        subject=f"{hostname} - {name}: {nr_failures} failures"
+        message=f"""
+{nr_failures} failures for {name} ({prev_failures} since last notification).
+Current status:
+
+{cur_status}
+
+
+Latest journal log:
+
+{journal_txt}
+"""
+
+    _send_email(subject, message, config)

sysalert/util.py

@@ -0,0 +1,92 @@
+
+import configparser
+import datetime
+import importlib
+import os
+import sqlite3
+import sys
+
+import systemd.journal
+
+import sysalert.db
+
+config_file='/etc/sysalert.ini'
+db_file='/var/lib/sysalert/sysalert.db'
+
+def _test_env():
+    # return True if all MONITOR environment variables are set, indicating
+    # utility is started as a handler for OnSuccess= or OnError=
+    return all([x in os.environ for x in (
+            'MONITOR_SERVICE_RESULT',
+            'MONITOR_EXIT_CODE',
+            'MONITOR_EXIT_STATUS',
+            'MONITOR_INVOCATION_ID',
+            'MONITOR_UNIT',
+            )])
+
+def register_exit(config, db):
+    service_name=sys.argv[1]
+    if service_name in config.sections():
+        section_name = service_name
+    else:
+        section_name = config.default_section
+    alert_method = config.get(section_name, 'alert_method')
+    alert_config = {x[0]:x[1] for x in config.items(alert_method)}
+    alert = importlib.import_module(alert_method)
+
+    if os.environ['MONITOR_SERVICE_RESULT'] == 'success':
+        # exit with success status
+        failures = sysalert.db.get_failures(service_name, db)
+        sysalert.db.register_success(service_name, db)
+        try:
+            do_alert = config.getboolean(section_name, 'recovery_alert')
+        except ValueError:
+            if config.get(section_name, 'recovery_alert') == 'if-alerted' and failures:
+                do_alert = True
+            else:
+                do_alert = False
+
+        if do_alert:
+            alert.success(service_name, failures, alert_config)
+        return 0
+
+    # exit with failed status
+    failures = sysalert.db.get_failures(service_name, db)
+    if len(failures) < config.getint(section_name, 'max_failures'):
+        sysalert.db.register_failure(service_name, None, db)
+        return 0
+
+    now = datetime.datetime.now()
+    last_alert = now
+    for f in reversed(failures):
+        if f['alert_method']:
+            last_alert = f['timestamp']
+            break
+
+    diff = datetime.timedelta(seconds=config.getint(section_name, 'resend_alert_time', 0))
+    if now != last_alert and last_alert + diff > now:
+        sysalert.db.register_failure(service_name, None, db)
+        return 0
+
+    # refresh failure list before alerting so we include this failure in alert
+    # call.
+    sysalert.db.register_failure(service_name, alert_method, db)
+    failures = sysalert.db.get_failures(service_name, db)
+    alert.failure(service_name, failures, alert_config)
+    return 0
+
+
+def cli():
+    config = configparser.ConfigParser()
+    config.read(config_file)
+    db = sysalert.db.init(db_file)
+
+    if _test_env():
+        # invoked by systemd
+        ret = register_exit(config, db)
+
+    sysalert.db.close(db)
+    return ret
+
+if __name__ == '__main__':
+  sys.exit(cli())