| Service name | Restart procedure |
|---|
| Ceilometer |
-
Log in to a controller node CLI.
-
Restart the Ceilometer services:
# service ceilometer-agent-central restart
# service ceilometer-api restart
# service ceilometer-agent-notification restart
# service ceilometer-collector status restart
-
Verify the status of the Ceilometer services. See Verify an OpenStack service status.
-
Repeat step 1 - 3 on all controller nodes.
|
| Cinder |
-
Log in to a controller node CLI.
-
Restart the Cinder services:
# service cinder-api restart
# service cinder-scheduler restart
-
Verify the status of the Cinder services. See Verify an OpenStack service status.
-
Repeat step 1 - 3 on all controller nodes.
-
On every node with Cinder role, run:
# service cinder-volume restart
# service cinder-backup restart
-
Verify the status of the cinder-volume and cinder-backup services.
|
| Corosync/Pacemaker |
-
Log in to a controller node CLI.
-
Restart the Corosync and Pacemaker services:
# service corosync restart
# service pacemaker restart
-
Verify the status of the Corosync and Pacemaker services. See Verify an OpenStack service status.
-
Repeat step 1 - 3 on all controller nodes.
|
| Glance |
-
Log in to a controller node CLI.
-
Restart the Glance services:
# service glance-api restart
# service glance-registry restart
-
Verify the status of the Glance services. See Verify an OpenStack service status.
-
Repeat step 1 - 3 on all controller nodes.
|
| Horizon |
Since the Horizon service is available through the Apache server, you should restart the Apache service on all controller nodes:
-
Log in to a controller node CLI.
-
Restart the Apache server:
# service apache2 restart
-
Verify whether the Apache service is successfully running after restart:
-
Verify whether the Apache ports are opened and listening:
# netstat -nltp | egrep apache2
-
Repeat step 1 - 3 on all controller nodes.
|
| Ironic |
-
Log in to a controller node CLI.
-
Restart the Ironic services:
# service ironic-api restart
# service ironic-conductor restart
-
Verify the status of the Ironic services. See Verify an OpenStack service status.
-
Repeat step 1 - 3 on all controller nodes.
-
On any controller node, run the following command for the nova-compute service configured to work with Ironic:
# crm resource restart p_nova_compute_ironic
-
Verify the status of the p_nova_compute_ironic service.
|
| Keystone |
Since the Keystone service is available through the Apache server, complete the following steps on all controller nodes:
-
Log in to a controller node CLI.
-
Restart the Apache server:
# service apache2 restart
-
Verify whether the Apache service is successfully running after restart:
-
Verify whether the Apache ports are opened and listening:
# netstat -nltp | egrep apache2
-
Repeat step 1 - 3 on all controller nodes.
|
| MySQL |
-
Log in to any controller node CLI.
-
Run the following command:
# pcs status | grep -A1 mysql
In the output, the resource clone_p_mysql should be in the Started status.
-
Disable the clone_p_mysql resource:
# pcs resource disable clone_p_mysqld
-
Verify that the resource clone_p_mysqld is in the Stopped status:
# pcs status | grep -A2 mysql
It may take some time for this resource to be stopped on all controller nodes.
-
Disable the clone_p_mysql resource:
# pcs resource enable clone_p_mysqld
-
Verify that the resource clone_p_mysqld is in the Started status again on all controller nodes:
# pcs status | grep -A2 mysql
Warning
Use the pcs commands instead of crm for restarting the service. The pcs tool correctly stops the service according to the quorum policy preventing MySQL failures.
|
| Neutron |
Use the following restart steps for the DHCP Neutron agent as an example for all Neutron agents.
-
Log in to any controller node CLI.
-
Verify the DHCP agent status:
# pcs resource show | grep -A1 neutron-dhcp-agent
The output should contain the list of all controllers in the Started status.
-
Stop the DHCP agent:
# pcs resource disable clone_neutron-dhcp-agent
-
Verify the Corosync status of the DHCP agent:
# pcs resource show | grep -A1 neutron-dhcp-agent
The output should contain the list of all controllers in the Stopped status.
-
Verify the neutron-dhcp-agent status on the OpenStack side:
The output table should contain the DHCP agents for every controller node with xxx in the alive column.
-
Start the DHCP agent on every controller node:
# pcs resource enable clone_neutron-dhcp-agent
-
Verify the DHCP agent status:
# pcs resource show | grep -A1 neutron-dhcp-agent
The output should contain the list of all controllers in the Started status.
-
Verify the neutron-dhcp-agent status on the OpenStack side:
The output table should contain the DHCP agents for every controller node with :-) in the alive column and True in the admin_state_up column.
|
| Nova |
-
Log in to a controller node CLI.
-
Restart the Nova services:
# service nova-api restart
# service nova-cert restart
# service nova-compute restart
# service nova-conductor restart
# service nova-consoleauth restart
# service nova-novncproxy restart
# service nova-scheduler restart
# service nova-spicehtml5proxy restart
# service nova-xenvncproxy restart
-
Verify the status of the Nova services. See Verify an OpenStack service status.
-
Repeat step 1 - 3 on all controller nodes.
-
On every compute node, run:
# service nova-compute restart
-
Verify the status of the nova-compute service.
|
| RabbitMQ |
-
Log in to any controller node CLI.
-
Disable the RabbitMQ service:
# pcs resource disable master_p_rabbitmq-server
-
Verify whether the service is stopped:
# pcs status | grep -A2 rabbitmq
-
Enable the service:
# pcs resource enable master_p_rabbitmq-server
During the startup process, the output of the pcs status command can show all existing RabbitMQ services in the Slaves mode.
-
Verify the service status:
# rabbitmqctl cluster_status
In the output, the running_nodes field should contain all controllers’ host names in the rabbit@<HOSTNAME> format. The partitions field should be empty.
|
| Swift |
-
Log in to a controller node CLI.
-
Restart the Swift services:
# service swift-account-auditor restart
# service swift-account restart
# service swift-account-reaper restart
# service swift-account-replicator restart
# service swift-container-auditor restart
# service swift-container restart
# service swift-container-reconciler restart
# service swift-container-replicator restart
# service swift-container-sync restart
# service swift-container-updater restart
# service swift-object-auditor restart
# service swift-object restart
# service swift-object-reconstructor restart
# service swift-object-replicator restart
# service swift-object-updater restart
# service swift-proxy restart
-
Verify the status of the Swift services. See Verify an OpenStack service status.
-
Repeat step 1 - 3 on all controller nodes.
|
Operating on more than one unit at a time
Systemd supports wildcarding for unit operations. To restart every service in devstack you can do that following:
sudo systemctl restart devstack@*
Or to see the status of all Nova processes you can do:
sudo systemctl status devstack@n-*
We’ll eventually make the unit names a bit more meaningful so that it’s easier to understand what you are restarting.
devstack 单元测试
devstack 重启服务