Administrative Procedures for RHEL High Availability Clusters - Starting All Nodes of a RHEL 7 Cluster
Contents
Overview
Applicable Environments
- Red Hat Enterprise Linux (RHEL) 7 with the High Availability Add-On
Situations Where This Procedure May Be Useful
- A cluster has just been created/configured, and is ready to be started
- All nodes of a cluster were taken down for maintenance or troubleshooting activities, and are ready to be started again
- Some nodes of a cluster are running, and the rest need to be started. (This procedure will not trigger any change directly in nodes already running).
What This Procedure Accomplishes
The nodes that are currently available and being operated on in this procedure have their pacemaker service started, causing the corosync service to start as a dependency. This will lead to the nodes attempting to communicate with each other, form a membership amongst any that are present, and begin participating in the cluster.
Once the nodes are started and form a membership with quorum, those members may be eligible to start hosting resources and participating in cluster-related activity.
If other members of the cluster are not present or able to communicate, the available nodes may have to wait for them before becoming operational, depending on the corosync configuration.
Procedure: Starting All Nodes of a RHEL 7 Cluster
Optional Task: Start and Authenticate pcsd
If pcs will be used to manage the state of the nodes as far as being started or stopped, enabled or disabled (on boot), then pcsd will need to be running.
If the node(s) where pcs commands will be executed from has not yet authenticated to each node's pcsd daemon - such as if this cluster was just created or any node was recently added - then authentication will need to be performed first.
Task: Start Cluster on All Nodes
If pcsd is running throughout the cluster, the simplest way to start all nodes is to execute the following command from one node:
# pcs cluster start --all
Alternatively, each node can be started individually, but ideally this should be run at approximately the same time to ensure they are all present to join the membership and do not keep any nodes waiting:
# pcs cluster start
If pcsd is not running, systemctl can be used on each node to start pacemaker:
# systemctl start pacemaker.service
Optional Task: Enable Cluster Services to Start On Boot
If the nodes of the cluster should automatically join the cluster on any subsequent boots, then enable the High Availability services throughout the cluster.