pgsql: fix failback operation refs #33389

repmgr v4.2 (on debian 10) has a bug, it doesn't detect role change after back to normal state: https://github.com/2ndQuadrant/repmgr/issues/524#issuecomment-459230305. The impact is after 1st successful switchover and failback repmgr will do no more switchover until restarted. This commit adds a restart operation to repmgrd (automatically executed on standby-to-primary tag).

ansible/playbooks/postgres-maintenance.yml

@@ -70,7 +70,7 @@
 
 - name: REPMGRD RESTART
   hosts: postgres
-  tags: [ 'never', 'restart-repmgrd' ]
+  tags: [ 'never', 'restart-repmgrd', 'standby-to-primary' ]
   tasks:
     - name: kill repmgrd
       command: "pkill repmgrd"

ansible/roles/postgres-ha/templates/repmgr-event.py.j2

-#!/usr/bin/env python3
-
-"""
-event_notification_command='/usr/local/bin/repmgr_event.py %n %e %s "%t" "%d"'
-"""
-
-import argparse
-import json
-import logging
-import logging.handlers
-import sys
-
-PARSER = argparse.ArgumentParser()
-PARSER.add_argument("node_id")
-PARSER.add_argument("event_type")
-PARSER.add_argument("success")
-PARSER.add_argument("timestamp")
-PARSER.add_argument("details")
-ARGS = PARSER.parse_args()
-
-MONIT_EVENTS = (
-    "standby_promote",
-    "repmgrd_failover_promote",
-    "repmgrd_failover_follow",
-)
-
-LOG_DATA = (ARGS.timestamp, ARGS.node_id, ARGS.event_type, ARGS.success, ARGS.details)
-
-
-def main():
-    """Log repmgrd events."""
-
-    # do not process unwanted events
-    if ARGS.event_type not in MONIT_EVENTS:
-        sys.exit(0)
-
-    # load config if exists
-    try:
-        with open("/etc/postgresql/{{ repmgr_pg_version }}/{{ repmgr_pg_cluster }}/repmgr-event.json") as data_file:
-            data = json.load(data_file)
-    except EnvironmentError:
-        data = {}
-
-    # get config, fallback on defaults
-    log_file = data.get("log", "/var/log/repmgr-event.log")
-    mail = data.get("mail", None)
-    if mail:
-        host = mail.get("host", "localhost")
-        sender = mail.get("sender", "root@localhost")
-        recipients = mail.get("recipients", ["root@localhost"])
-        credentials = mail.get("credentials", None)
-        secure = mail.get("secure", None)
-    # from args
-    subject = "repmgr event: {}".format(ARGS.event_type)
-
-    # logging conf
-    logger = logging.getLogger("repmgr_event")
-    logger.setLevel(logging.INFO)
-
-    # log to stream
-    stream_handler = logging.StreamHandler(sys.stdout)
-    logger.addHandler(stream_handler)
-
-    # log to file
-    file_handler = logging.handlers.RotatingFileHandler(
-        log_file, maxBytes=102400, backupCount=7
-    )
-    logger.addHandler(file_handler)
-
-    # log to mail
-    if mail:
-        mail_handler = logging.handlers.SMTPHandler(
-            host, sender, recipients, subject, credentials, secure
-        )
-        logger.addHandler(mail_handler)
-
-    # write log
-    logger.info("%s - %s:%s:%s - %s", *LOG_DATA)
-
-    sys.exit(0)
-
-
-if __name__ == "__main__":
-    main()