Configuring EVMS for a High-Availability Cluster
- Installing Linux-HA
In order to run EVMS in a high-availability cluster environment, you must first
have the Linux-HA software installed
and configured. If you do not already have Linux-HA installed, you should
have downloaded the necessary packages in a previous
These instructions will not cover the full details of installing and setting
up Linux-HA. Please read through the the Linux-HA
Started and FAQ
'n Tips documents for in-depth instructions on installing the base
If you downloaded the RPM packages for Linux-HA, they can easily be installed
with the following command.
rpm -i heartbeat-pils-xxx.arch.rpm heartbeat-stonith-xxx.arch.rpm \
where xxx is the version of heartbeat and arch is the
architecture for which the RPMs were built.
If you are running on a two-node cluster, it is important to configure the
STONITH package, which will provide the fail-over capabilities for the cluster.
EVMS will still operate without STONITH configured, but the integrity of the
data on shared volumes may be at risk. EVMS's ability to reassign ownership
of shared disks is limited to the support provided by the cluster manager
(Linux-HA). Please refer to the STONITH device manual for instructions on how
to configure the STONITH device.
- Configuring EVMS for Linux-HA
Perform the following steps on all the cluster nodes.
- Configure CCM Services
EVMS depends on CCM services offered by Heartbeat. Execute the following
command to configure CCM.
echo "respawn haclient /usr/lib/heartbeat/ccm" >> /etc/ha.d/ha.cf
- Create Communication Channels
Linux-HA expects its clients to setup a private communication channel. Create
the following fifos.
chown root:haclient /var/lib/heartbeat/api/evms.req
chown root:haclient /var/lib/heartbeat/api/evms.rsp
chmod 400 /var/lib/heartbeat/api/evms.req
chmod 600 /var/lib/heartbeat/api/evms.rsp
- Tell Linux-HA to Activate EVMS
To ensure the EVMS daemon is activated whenever Linux-HA starts, execute the
echo "respawn root /sbin/evmsd" >> /etc/ha.d/ha.cf
This command assumes that the evmsd daemon is installed in
/sbin. If you installed EVMS in a non-default location, change the
above command to refer to the correct location of evmsd.
- Configuring EVMS for Fail-Over
Ensure that evms_activate is run before heartbeat at system
start-up. If Linux-HA starts before EVMS, the fail-overs may not work correctly.
Linux-HA currently supports fail-over only on two-node clusters. Clusters
with more than two nodes cannot be configured for failover.
Only private disk-groups (owned by a single node in the cluster) can be failed
over. Shared disk-groups are available to all nodes in the cluster
simultaneously, and thus their ownership will not change when one node fails.
- For each private disk-group that you wish to participate in fail-overs, add
an entry to the /etc/ha.d/haresources file.
For example, if disk-groups dg1 and dg2 are currently owned by node n1, and
should be failed-over together when n1 dies, then add the following entry.
n1 evms_failover::dg1 evms_failover::dg2
For more details about the semantics of Linux-HA resource groups, please see
Getting Started guide.
- Validate that the /etc/ha.d/ha.cf and /etc/haresources
files are the same on all nodes of the cluster.
- Restart the Linux-HA cluster manager on all nodes by executing the