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

#!/bin/bash

RETRIES="{{ handler_health_rgw_check_retries }}"
DELAY="{{ handler_health_rgw_check_delay }}"
HOST_NAME="{{ ansible_facts['hostname'] }}"
RGW_NUMS={{ rgw_instances | length | int }}
RGW_FRONTEND_SSL_CERT={{ radosgw_frontend_ssl_certificate }}
if [ -n "$RGW_FRONTEND_SSL_CERT" ]; then
    RGW_PROTOCOL=https
else
    RGW_PROTOCOL=http
fi
INSTANCES_NAME=({% for i in rgw_instances %}{{ i.instance_name }} {% endfor %})
RGW_IPS=({% for i in rgw_instances %}{{ i.radosgw_address }} {% endfor %})
RGW_PORTS=({% for i in rgw_instances %}{{ i.radosgw_frontend_port }} {% endfor %})
declare -a DOCKER_EXECS
declare -a SOCKET_PREFIX
for ((i=0; i<${RGW_NUMS}; i++)); do
  SOCKET_PREFIX[i]="/var/run/ceph/{{ cluster }}-client.rgw.${HOST_NAME}.${INSTANCES_NAME[i]}"
  DOCKER_EXECS[i]=""
{% if containerized_deployment | bool %}
  DOCKER_EXECS[i]="{{ container_binary }} exec ceph-rgw-${HOST_NAME}-${INSTANCES_NAME[i]}"
{% endif %}
done

check_socket() {
  local i=$1
  local succ=0
  local count=10
  # Wait and ensure the socket exists after restarting the daemon
  while [ $count -ne 0 ]; do
    SOCKET=$(grep ${SOCKET_PREFIX[i]} /proc/net/unix | awk '{ print $8 }')
    if [ -n "${SOCKET}" ]; then
      ${DOCKER_EXECS[i]} test -S ${SOCKET} && succ=$((succ+1)) && break
    fi
    sleep $DELAY
    let count=count-1
  done
  if [ $succ -ne 1 ]; then
    echo "Socket file ${SOCKET} could not be found, which means Rados Gateway is not running. Showing ceph-rgw unit logs now:"
    journalctl -u ceph-radosgw@rgw.${HOST_NAME}.${INSTANCES_NAME[i]}
    exit 1
  fi
}

check_for_curl_or_wget() {
  local i=$1
  if ${DOCKER_EXECS[i]} command -v wget &>/dev/null; then
    rgw_test_command="wget --no-check-certificate --tries 1 --quiet -O /dev/null"
  elif ${DOCKER_EXECS[i]} command -v curl &>/dev/null; then
    rgw_test_command="curl {{ '-g' if ip_version == 'ipv6' else '' }} -k --fail --silent --output /dev/null"
  else
    echo "It seems that neither curl nor wget are available on your system."
    echo "Cannot test rgw connection."
    exit 0
  fi
}

check_rest() {
  local i=$1
  check_for_curl_or_wget ${i}
  local succ=0
  while [ $RETRIES -ne 0 ]; do
    ${DOCKER_EXECS[i]} $rgw_test_command $RGW_PROTOCOL://${RGW_IPS[i]}:${RGW_PORTS[i]} && succ=$((succ+1)) && break
    sleep $DELAY
    let RETRIES=RETRIES-1
  done
  if [ $succ -ne 1 ]; then
    # If we reach this point, it means there is a problem with the connection to rgw
    echo "Error connecting locally to Rados Gateway service: $RGW_PROTOCOL://${RGW_IPS[i]}:${RGW_PORTS[i]}"
    exit 1
  fi
}

for ((i=0; i<${RGW_NUMS}; i++)); do
  # First, restart the daemon

  # Check if systemd unit exists
  # This is needed for new instances are the restart might trigger before the deployment
  if systemctl list-units --full --all | grep -q "ceph-radosgw@rgw.${HOST_NAME}.${INSTANCES_NAME[i]}"; then
    systemctl restart ceph-radosgw@rgw.${HOST_NAME}.${INSTANCES_NAME[i]}
  else
    echo "Systemd unit ceph-radosgw@rgw.${HOST_NAME}.${INSTANCES_NAME[i]} does not exist."
    continue
  fi

  # Check socket files
  check_socket ${i}
  # Check rest
  check_rest ${i}
done