cms_failconf(1m)cms_failconf(1m)NAMEcms_failconf - CXFS cluster node failure configuration
SYNOPSIScms_failconf [ option ... ]
DESCRIPTION
The cms_failconf utility is used to configure the action taken by the
surviving nodes when a CXFS node fails (i.e. falls out of CXFS cluster
membership). Note that users would normally configure fail actions (as
part of initial node definition) via the CXFS GUI or cmgr command, as
described in the CXFS Version 2 Software Installation and Administration
Guide. Use of cms_failconf is restricted to users with the
CAP_SYSINFO_MGT capability.
The -q option cms_failconf displays the current failure action
configuration for all nodes currently in the CXFS cluster. The -s option
can be used to change the failure action hierarchy for a particular node.
The syntax for this is
cms_failconf-s node:[fence|fencereset][,reset]shutdown
i.e. optionally either fence or fencereset and if that fails, then
optionally reset. In any case, the final action is shutdown, meaning the
failed node will initiate CXFS shutdown, withdraw from cluster
membership, unmount it's CXFS filesystems, automatically attempt to
rejoin membership and then mount it's CXFS filesystems again.
The default failure action hierarchy for all nodes is reset, shutdown.
This means the surviving nodes will attempt to reset the failed node
using the serial reset lines, if configured. Consult the CXFS Software
Installation and Administration Guide for instructions on configuring
reset lines.
The fence failure action means the surviving nodes will attempt to
disable the Fibre Channel port in the Fibre Channel switch for the failed
node, thus preventing it from performing any further I/O. The failed
node will either need to be rebooted, or may, if sufficiently healthy,
recover from fencing and re-apply to join the cluster membership (the
recovery process is automatic). The surviving nodes will lower the fence
for the failed node when it rejoins membership, thus allowing the node to
mount CXFS filesystems again.
The fencereset fail action is similar to fence, but specifies to also
reset the node immediately after raising the fence rather than allowing
it to attempt recovery. If raising the fence fails, the reset is not
attempted (unless the next fail action is reset).
For further details and instructions for setting up reset lines and how
to configure a cluster to enable fencing, see the CXFS Software
Installation and Administration Guide.
Page 1
cms_failconf(1m)cms_failconf(1m)SEE ALSOhafence(1m), and the CXFS Software Installation and Administration Guide.
Page 2
