onnode — run commands on CTDB cluster nodes
onnode is a utility to run commands on a specific node of a CTDB cluster, or on all nodes.
NODES specifies which node(s) to run
a command on. See section NODES
SPECIFICATION for details.
COMMAND can be any shell command. The
onnode utility uses ssh or rsh to connect to the remote nodes
and run the command.
Execute COMMAND in the current working directory on the specified nodes.
Specify an alternative nodes FILENAME to use instead of
the default. This option overrides the CTDB_NODES_FILE
and CTDB_NODES variables. See the discussion of
/usr/local/etc/ctdb/nodes in the FILES section
for more details.
Keep standard input open, allowing data to be piped to
onnode. Normally onnode closes stdin to avoid surprises
when scripting. Note that this option is ignored when
-p or if
SSH is set
to anything other than "ssh".
Allow nodes to be specified by name rather than node
numbers. These nodes don't need to be listed in the nodes
file. You can avoid the nodes file entirely by combining
Causes standard output from each node to be saved into a
file with name PREFIX.
Run COMMAND in parallel on the specified nodes. The default is to run COMMAND sequentially on each node.
Push files to nodes. Names of files to push are specified rather than the usual command. Quoting is fragile/broken - filenames with whitespace in them are not supported.
Do not print node addresses. Normally, onnode prints informational node addresses if more than one node is specified. This overrides -v.
Print node addresses even if only one node is specified. Normally, onnode prints informational node addresses when more than one node is specified.
Show a short usage guide.
Nodes can be specified via numeric node numbers (from 0 to N-1) or mnemonics. Multiple nodes are specified using lists of nodes, separated by commas, and ranges of numeric node numbers, separated by dashes. If nodes are specified multiple times then the command will be executed multiple times on those nodes. The order of nodes is significant.
The following mnemonics are available:
A node where ctdbd is running. This semi-random but there is a bias towards choosing a low numbered node.
All nodes that are not disconnected, banned, disabled or unhealthy.
All nodes that are not disconnected.
The following command would show the process ID of ctdbd on all nodes
onnode all ctdb getpid
The following command would show the last 5 lines of log on each node, preceded by the node's hostname
onnode all "hostname; tail -5 /usr/local/var/log/log.ctdb"
The following command would restart the ctdb service on all nodes, in parallel.
onnode -p all service ctdb restart
The following command would run ./foo in the current working directory, in parallel, on nodes 0, 2, 3 and 4.
onnode -c -p 0,2-4 ./foo
Directory containing CTDB configuration files. The
Name of alternative nodes file to use instead of the default. See the FILES section for more details.
Default file containing a list of each node's IP address or hostname.
As above, a file specified via the
CTDB_NODES_FILE is given precedence. If a
relative path is specified and no corresponding file
exists relative to the current directory then the file is
also searched for in the
CTDB_NODES_FILE is not set and
CTDB_NODES is set in configuration then the
file pointed to by
CTDB_NODES is used.
Otherwise the default is
CTDB_BASE defaults to
If this file exists it is sourced by onnode. The main
purpose is to allow the administrator to set
SSH to something other than "ssh". In this
case the -t option is ignored. For example, the
administrator may choose to use use rsh instead of ssh.