Manual Pages
Table of Contents
na_cf - controls the takeover and giveback operations of
the filers in a cluster
cf [ disable | enable | forcegiveback | forcetakeover [
-df ] | giveback [ -f ] | hw_assist [ status | test
stats [ clear ] ] | monitor | partner | status [ -t ]
takeover [ -f ] | [ -n ]]
cf nfo [ enable | disable ] disk_shelf
cf nfo status
The cf command controls the cluster failover monitor,
which determine when takeover and giveback operations take
place within a cluster.
The cf command is available only if your filer has the
cluster license.
-
disable
- Disables the takeover capability of
both filers in the cluster.
-
enable
- Enables the takeover capability of
both filers in the cluster.
-
forcegiveback
- forcegiveback is dangerous and can
lead to data corruption; in almost all
cases, use cf giveback -f instead.
Forces the live filer to give back the
resources of the failed filer even
though the live filer determines that
doing so might result in data corruption
or cause other severe problems.
giveback will refuse to giveback under
these conditions. Using the forcegiveback
option forces a giveback. When
the failed filer reboots as a result
of a forced giveback, it displays the
following message:
partner giveback incomplete, some data
may be lost
forcetakeover [-f] forcetakeover is dangerous and can
lead to data corruption; in almost all
cases, use cf takeover instead.
Forces one filer to take over its
partner even though the filer detects
an error that would otherwise prevent
a takeover. For example, normally, if
a detached or faulty ServerNet cable
between the filers causes the filers'
NVRAM contents to be unsynchronized,
takeover is disabled. However, if you
enter the cf forcetakeover command,
the filer takes over its partner
despite the unsynchronized NVRAM contents.
This command might cause the
filer being taken over to lose client
data. If you use the -f option, the
cf command allows such a forcetakeover
to proceed without requiring confirmation
by the operator.
forcetakeover -d[f] Forces a filer to take over its partner
in all cases where a forcetakeover
would fail. In addition it will force
a takeover even if some partner mailbox
disks are inaccessible. It can
only be used when cluster_remote is
licensed.
forcetakeover -d is very dangerous.
Not only can it cause data corruption,
if not used carefully, it can also
lead to a situation where both the
filer and it's partner are operational
(split brain). As such, it should only
be used as a means of last resort when
the takeover and forcetakeover commands
are unsuccessful in achieving a
takeover. The operator must ensure
that the partner filer does not become
operational at any time while a filer
is in a takeover mode initiated by the
use of this command. In conjunction
with RAID mirroring, it can allow
recovery from a disaster when the two
filers in the cluster are located at
two distant sites. The use of -f
option allows this command to proceed
without requiring confirmation by the
operator.
-
giveback [ -f ]
- Initiates a giveback of partner
resources. Once the giveback is complete,
the automatic takeover capability
is disabled until the partner is
rebooted. A giveback fails if outstanding
CIFS sessions, active system
dump processes, or other filer operations
makes a giveback dangerous or
disruptive. If you use the -f option,
the cf command allows such a giveback
to proceed as long as it would not
result in data corruption or filer
error.
hw_assist [ status | test | stats [ clear ] ]
Displays information related to the
hardware-assisted takeover functionality.
Use the cf hw_assist status command
to display the hardware-assisted
functionality status of the local as
well as the partner filer. If hardware-assisted
status is inactive, the
command displays the reason and if
possible, a corrective action. Use
the cf hw_assist test command to validate
the hardware-assisted takeover
configuration. An error message is
printed if hardware-assisted takeover
configuration can not be validated.
Use the cf hw_assist stats command to
display the statistics for all
hw_assist alerts received by the
filer. Use cf hw_assist stats clear to
clear hardware-assisted functionality
statistics.
-
monitor
- Displays the time, the state of the
local filer and the time spent in this
state, the host name of the partner
and the state of cluster failover monitor
(whether enabled or disabled).
If the partner has not been taken over
currently, the status of the partner
and that of the interconnect are displayed
and any ongoing giveback or
scheduled takeover operations are
reported.
-
partner
- Displays the host name of the partner.
If the name is unknown, the cf command
displays ``partner.''
-
status
- Displays the current status of the
local filer and the cluster. If you
use the -t option, displays the status
of the node as time master or slave.
takeover [ -f ] | [ -n ]
Initiates a takeover of the partner.
If you use the -f option, the cf command
allows such a takeover to proceed
even if it will abort a coredump on
the other filer.
If you use the -n option, the cf command
allows a takeover to proceed even
if the partner node was running an
incompatible version of Data ONTAP.
The partner node must be cleanly
halted in order for this option to
succeed. This is used as part of a
nondisruptive upgrade process.
nfo [ enable | disable ] disk_shelf
Enables or disables negotiated
failover on disk shelf count mismatch.
This command is obsolete. Option
cf.takeover.on_disk_shelf_miscompare
replaces it.
Negotiated failover is a general
facility which supports negotiated
failover on the basis of decisions
made by various modules. disk_shelf
is the only negotiated failover module
currently implemented. When communication
is first established over the
interconnect between the local filer
and its partner, a list of disk
shelves seen by each node on its A and
B loops is exchanged. If a filer sees
that the count of shelves that the
partner sees on its B loops is greater
than the filer's count of shelves on
its A loops, the filer concludes that
it is ``impaired'' (as it sees fewer
of its shelves than its partner does)
and asks the partner to take it over.
If the partner is not itself impaired,
it will accept the takeover request
and, in turn, ask the requesting filer
to shut down gracefully. The partner
takes over after the requesting node
shuts down, or after a time-out period
of approximately 3 minutes expires.
The comparison of disk shelves is only
done when communication between the
filers is established or re-established
(for example, after a node
reboots).
-
nfo status
- Displays the current negotiated
failover status.
This command is obsolete. Use cf status
instead.
na_partner(1)
Table of Contents