lvm2/man/clvmd.8_main

.TH CLVMD 8 "LVM TOOLS #VERSION#" "Red Hat Inc" \" -*- nroff -*-
.
.SH NAME
.
clvmd \(em cluster LVM daemon
.
.SH SYNOPSIS
.
.ad l
.B clvmd
.RB [ -C ]
.RB [ -d
.RI [ value ]]
.RB [ -E
.IR lock_uuid ]
.RB [ -f ]
.RB [ -h ]
.RB [ -I
.IR cluster_manager ]
.RB [ -R ]
.RB [ -S ]
.RB [ -t
.IR timeout ]
.RB [ -T
.IR start_timeout ]
.RB [ -V ]
.ad b
.
.SH DESCRIPTION
.
clvmd is the daemon that distributes LVM metadata updates around a cluster.
It must be running on all nodes in the cluster and will give an error
if a node in the cluster does not have this daemon running.

Also see \fBlvmlockd\fP(8) for a newer method of using LVM on shared
storage.
.
.SH OPTIONS
.
.HP
.BR -C
.br
Only valid if \fB-d\fP is also specified.
Tells all clvmds in a cluster to enable/disable debug logging.
Without this switch, only the local clvmd will change its debug level to that
given with \fB-d\fP.
.br
This does not work correctly if specified on the command-line that starts clvmd.
If you want to start clvmd \fBand\fP
enable cluster-wide logging then the command needs to be issued twice, eg:
.br
.BR clvmd
.br
.BR clvmd\ -d2
.
.HP
.BR -d
.RI [ value ]
.br
Set debug logging level.
If \fB-d\fP is specified without a \fIvalue\fP
then 1 is assumed. \fIValue\fP can be:
.PD 0
.IP
.BR 0
\(em Disabled
.IP
.BR 1
\(em Sends debug logs to stderr (implies \fB-f\fP)
.IP
.BR 2
\(em Sends debug logs to \fBsyslog\fP(3)
.PD
.
.HP
.BR -E
.IR lock_uuid
.br
Pass lock uuid to be reacquired exclusively when clvmd is restarted.
.
.HP
.BR -f
.br
Don't fork, run in the foreground.
.
.HP
.BR -h
.br
Show help information.
.
.HP
.BR -I
.IR cluster_manager
.br
Selects the cluster manager to use for locking and internal
communications. As it is quite possible to have multiple managers available on
the same system you might have to manually specify this option to override the
search.

By default, omit \fB-I\fP is equivalent to \fB-Iauto\fP.
Clvmd will use the first cluster manager that succeeds,
and it checks them in a predefined order
.BR cman ,
.BR corosync ,
.BR openais .
The available managers will be listed by order as part of the
\fBclvmd -h\fP output.
.
.HP
.BR -R
.br
Tells all the running instance of \fBclvmd\fP in the cluster to reload their device cache and
re-read the lvm configuration file \fBlvm.conf\fP(5). This command should be run whenever the
devices on a cluster system are changed.
.
.HP
.BR -S
.br
Tells the running \fBclvmd\fP to exit and reexecute itself, for example at the
end of a package upgrade. The new instance is instructed to reacquire
any locks in the same state as they were previously held. (Alternative
methods of restarting the daemon have the side effect of changing
exclusive LV locks into shared locks.)
.
.HP
.BR -t
.IR timeout
.br
Specifies the \fItimeout\fP for commands to run around the cluster. This should not
be so small that commands with many disk updates to do will fail, so you
may need to increase this on systems with very large disk farms.
The default is 60 seconds.
.
.HP
.BR -T
.IR start_timeout
.br
Specifies the start timeout for \fBclvmd\fP daemon startup. If the
daemon does not report that it has started up within this time then the parent
command will exit with status of 5. This does NOT mean that \fBclvmd\fP has
not started! What it means is that the startup has been delayed for some
reason; the most likely cause of this is an inquorate cluster though it
could be due to locking latencies on a cluster with large numbers of logical
volumes. If you get the return code of 5 it is usually not necessary to
restart \fBclvmd\fP it will start as soon as that blockage has cleared.
This flag is to allow startup scripts to exit in a timely fashion even if the
cluster is stalled for some reason.

The default is \fB0\fP (no timeout) and the value is in seconds. Don't set this too
small or you will experience spurious errors. 10 or 20 seconds might be
sensible.

This timeout will be ignored if you start \fBclvmd\fP with the \fB-d\fP.
.
.HP
.BR -V
.br
Display the version of the cluster LVM daemon.
.
.SH NOTES
.
.SS Activation
.
In a clustered VG, clvmd is used for activation, and the following values are
possible with \fBlvchange/vgchange -a\fP:
.IP \fBy\fP|\fBsy\fP
clvmd activates the LV in shared mode (with a shared lock),
allowing multiple nodes to activate the LV concurrently.
If the LV type prohibits shared access, such as an LV with a snapshot,
an exclusive lock is automatically used instead.
clvmd attempts to activate the LV concurrently on all nodes.
.IP \fBey\fP
clvmd activates the LV in exclusive mode (with an exclusive lock),
allowing a single node to activate the LV.
clvmd attempts to activate the LV concurrently on all nodes, but only
one will succeed.
.IP \fBly\fP
clvmd attempts to activate the LV only on the local node.
If the LV type allows concurrent access, then shared mode is used,
otherwise exclusive.
.IP \fBn\fP
clvmd attempts to deactivate the LV on all nodes.
.IP \fBln\fP
clvmd attempts to deactivate the LV on the local node.
.
.SH ENVIRONMENT VARIABLES
.TP
.B LVM_CLVMD_BINARY
The CLVMD binary to use when \fBclvmd\fP restart is requested.
Defaults to \fI#CLVMD_PATH#\fP.
.TP
.B LVM_BINARY
The LVM2 binary to use.
Defaults to \fI#LVM_PATH#\fP.
.SH FILES
.I #CLVMD_PATH#
.br
.I #LVM_PATH#
.SH SEE ALSO
.BR syslog (3),
.BR lvm.conf (5),
.BR lvm (8),
.BR lvmlockd (8),
.BR lvmsystemid (7)