Parallel devices¶

Some devices, like dual power or dual path, must all be turned off in parallel for fencing to succeed. To define multiple devices as being parallel to each other, use the same base dev_name with different suffixes and a colon separator between base name and suffix.

Format:

device dev_name:1 agent [args]
connect dev_name:1 node=nodeid [args]
connect dev_name:1 node=nodeid [args]
connect dev_name:1 node=nodeid [args]

device dev_name:2 agent [args]
connect dev_name:2 node=nodeid [args]
connect dev_name:2 node=nodeid [args]
connect dev_name:2 node=nodeid [args]

Example:

device foo:1 fence_foo ipaddr=1.1.1.1 login=x password=y
connect foo:1 node=1 port=1
connect foo:2 node=2 port=2
connect foo:3 node=3 port=3

device foo:2 fence_foo ipaddr=5.5.5.5 login=x password=y
connect foo:2 node=1 port=1
connect foo:2 node=2 port=2
connect foo:2 node=3 port=3

Unfencing¶

A node may sometimes need to "unfence" itself when starting. The unfencing command reverses the effect of a previous fencing operation against it. An example would be fencing that disables a port on a SAN switch. A node could use unfencing to re-enable its switch port when starting up after rebooting. (Care must be taken to ensure it's safe for a node to unfence itself. A node often needs to be cleanly rebooted before unfencing itself.)

To specify that a node should unfence itself for a given device, the unfence line is added after the connect lines.

Format:

device dev_name agent [args]
connect dev_name node=nodeid [args]
connect dev_name node=nodeid [args]
connect dev_name node=nodeid [args]
unfence dev_name

Example:

device foo fence_foo ipaddr=1.1.1.1 login=x password=y
connect foo node=1 port=1
connect foo node=2 port=2
connect foo node=3 port=3
unfence foo

Simple devices¶

In some cases, a single fence device is used for all nodes, and it requires no node-specific args. This would typically be a "bridge" fence device in which an agent is passing a fence request to another subsystem to handle. (Note that a "node=nodeid" arg is always automatically included in agent args, so a node-specific nodeid is always present to minimally identify the victim.)

In such a case, a simplified, single-line fence configuration is possible, with format:

fence_all agent [args]

Example:

fence_all dlm_stonith

A fence_all configuration is not compatible with a fence device configuration (above).

Unfencing can optionally be applied with:

fence_all agent [args]
unfence_all

Disabling resource directory¶

Lockspaces usually use a resource directory to keep track of which node is the master of each resource. The dlm can operate without the resource directory, though, by statically assigning the master of a resource using a hash of the resource name. To enable, set the per-lockspace nodir option to 1.

Example:

lockspace foo nodir=1

Lock-server configuration¶

The nodir setting can be combined with node weights to create a configuration where select node(s) are the master of all resources/locks. These master nodes can be viewed as "lock servers" for the other nodes.

Example of nodeid 1 as master of all resources:

lockspace foo nodir=1
master foo node=1

Example of nodeid's 1 and 2 as masters of all resources:

lockspace foo nodir=1
master foo node=1
master foo node=2

Lock management will be partitioned among the available masters. There can be any number of masters defined. The designated master nodes will master all resources/locks (according to the resource name hash). When no masters are members of the lockspace, then the nodes revert to the common fully-distributed configuration. Recovery is faster, with little disruption, when a non-master node joins/leaves.

There is no special mode in the dlm for this lock server configuration, it's just a natural consequence of combining the "nodir" option with node weights. When a lockspace has master nodes defined, the master has a default weight of 1 and all non-master nodes have weight of 0. An explicit non-zero weight can also be assigned to master nodes, e.g.

lockspace foo nodir=1
master foo node=1 weight=2
master foo node=2 weight=1

In which case node 1 will master 2/3 of the total resources and node 2 will master the other 1/3.