ceph-ansible/roles/ceph-handler/templates/restart_mgr_daemon.sh.j2

#!/bin/bash

RETRIES="{{ handler_health_mgr_check_retries }}"
DELAY="{{ handler_health_mgr_check_delay }}"
MGR_NAME="{{ ansible_hostname }}"
{% if containerized_deployment | bool %}
DOCKER_EXEC="{{ container_binary }} exec ceph-mgr-{{ ansible_hostname }}"
{% endif %}

# Backward compatibility
$DOCKER_EXEC test -S /var/run/ceph/{{ cluster }}-mgr.{{ ansible_fqdn }}.asok && SOCKET=/var/run/ceph/{{ cluster }}-mgr.{{ ansible_fqdn }}.asok
$DOCKER_EXEC test -S /var/run/ceph/{{ cluster }}-mgr.{{ ansible_hostname }}.asok && SOCKET=/var/run/ceph/{{ cluster }}-mgr.{{ ansible_hostname }}.asok

systemctl reset-failed ceph-mgr@${MGR_NAME}
# First, restart the daemon
systemctl restart ceph-mgr@${MGR_NAME}

# Wait and ensure the socket exists after restarting the daemds
while [ $RETRIES -ne 0 ]; do
  $DOCKER_EXEC test -S $SOCKET && exit 0
  sleep $DELAY
  let RETRIES=RETRIES-1
done
# If we reach this point, it means the socket is not present.
echo "Socket file ${SOCKET} could not be found, which means ceph manager is not running. Showing ceph-mgr unit logs now:"
journalctl -u ceph-mgr@${MGR_NAME}
exit 1