Installation guide

The advantage of using the optimized consensus timeout for two-node clusters is that overall failover
time is reduced for the two-node case, since consensus is not a function of the token timeout.
Note that for two-node autodetection in cman, the number of physical nodes is what matters and not
the presence of the two_no d e= 1 directive in the cl uster. co nf file.
7.3. Configuring Fencing
Configuring fencing consists of (a) specifying one or more fence devices in a cluster and (b)
specifying one or more fence methods for each node (using a fence device or fence devices
specified).
Note
It is recommended that you configure multiple fencing mechanisms for each node. A fencing
device can fail due to network split, a power outage, or a problem in the fencing device itself.
Configuring multiple fencing mechanisms can reduce the likelihood that the failure of a
fencing device will have fatal results.
Based on the type of fence devices and fence methods required for your configuration, configure
cl uster.co nf as follows:
1. In the fencedevi ces section, specify each fence device, using a fenced evi ce element
and fence-device dependent attributes. Example 7.3, “ APC Fence Device Added to
cl uster.co nf ” shows an example of a configuration file with an APC fence device added
to it.
2. At the clusterno d es section, within the fence element of each cl usternod e section,
specify each fence method of the node. Specify the fence method name, using the method
attribute, name. Specify the fence device for each fence method, using the d evi ce element
and its attributes, name and fence-device-specific parameters. Example 7.4, “ Fence Methods
Added to cl uster. co nf ” shows an example of a fence method with one fence device for
each node in the cluster.
3. For non-power fence methods (that is, SAN/storage fencing), at the cl usternod es section,
add an unfence section. This ensures that a fenced node is not re-enabled until the node
has been rebooted. When you configure a device that requires unfencing, the cluster must
first be stopped and the full configuration including devices and unfencing must be added
before the cluster is started. For more information about unfencing a node, refer to the
fence_no d e(8) man page.
The unfence section does not contain metho d sections like the fence section does. It
contains d evi ce references directly, which mirror the corresponding device sections for
fence, with the notable addition of the explicit action (actio n) of "on" or "enable". The
same fenced evi ce is referenced by both fence and unfence d evi ce lines, and the same
per-node arguments should be repeated.
Specifying the acti o n attribute as " on" or "enable" enables the node when rebooted.
Example 7.4, “Fence Methods Added to cl uster. co nf ” and Example 7.5, “ cl uster. co nf:
Multiple Fence Methods per Node” include examples of the unfence elements and attributed.
For more information about unfence refer to the fence_no d e man page.
4. Update the confi g _versio n attribute by incrementing its value (for example, changing
from co nfi g _versi o n= "2" to co nfi g _versio n= "3">).
T he con sensus Valu e for t ot em in a T wo- No de Clust er
109