NAME
ctdb - clustered tdb database management utility
SYNOPSIS
ctdb [ OPTIONS ] COMMAND ...
ctdb [-n <node>] [-Y] [-t <timeout>] [-T <timelimit>] [-? --help]
[--usage] [-d --debug=<INTEGER>] [--socket=<filename>]
DESCRIPTION
ctdb is a utility to view and manage a ctdb cluster.
OPTIONS
-n <pnn>
This specifies the physical node number on which to execute the
command. Default is to run the command on the deamon running on the
local host.
The physical node number is an integer that describes the node in
the cluster. The first node has physical node number 0.
-Y
Produce output in machine readable form for easier parsing by
scripts. Not all commands support this option.
-t <timeout>
How long should ctdb wait for the local ctdb daemon to respond to a
command before timing out. Default is 3 seconds.
-T <timelimit>
A limit on how long the ctdb command will run for before it will be
aborted. When this timelimit has been exceeded the ctdb command
will terminate.
-? --help
Print some help text to the screen.
--usage
Print useage information to the screen.
-d --debug=<debuglevel>
Change the debug level for the command. Default is 0.
--socket=<filename>
Specify the socketname to use when connecting to the local ctdb
daemon. The default is /tmp/ctdb.socket .
You only need to specify this parameter if you run multiple ctdb
daemons on the same physical host and thus can not use the default
name for the domain socket.
ADMINISTRATIVE COMMANDS
These are commands used to monitor and administrate a CTDB cluster.
pnn
This command displays the pnn of the current node.
status
This command shows the current status of the ctdb node.
node status
Node status reflects the current status of the node. There are
five possible states:
OK - This node is fully functional.
DISCONNECTED - This node could not be connected through the
network and is currently not participating in the cluster. If
there is a public IP address associated with this node it should
have been taken over by a different node. No services are
running on this node.
DISABLED - This node has been administratively disabled. This
node is still functional and participates in the CTDB cluster
but its IP addresses have been taken over by a different node
and no services are currently being hosted.
UNHEALTHY - A service provided by this node is malfunctioning
and should be investigated. The CTDB daemon itself is
operational and participates in the cluster. Its public IP
address has been taken over by a different node and no services
are currnetly being hosted. All unhealthy nodes should be
investigated and require an administrative action to rectify.
BANNED - This node failed too many recovery attempts and has
been banned from participating in the cluster for a period of
RecoveryBanPeriod seconds. Any public IP address has been taken
over by other nodes. This node does not provide any services.
All banned nodes should be investigated and require an
administrative action to rectify. This node does not perticipate
in the CTDB cluster but can still be communicated with. I.e.
ctdb commands can be sent to it.
STOPPED - A node that is stopped does not host any public ip
addresses, nor is it part of the VNNMAP. A stopped node can not
become LVSMASTER, RECMASTER or NATGW. This node does not
perticipate in the CTDB cluster but can still be communicated
with. I.e. ctdb commands can be sent to it.
generation
The generation id is a number that indicates the current
generation of a cluster instance. Each time a cluster goes
through a reconfiguration or a recovery its generation id will
be changed.
This number does not have any particular meaning other than to
keep track of when a cluster has gone through a recovery. It is
a random number that represents the current instance of a ctdb
cluster and its databases. CTDBD uses this number internally to
be able to tell when commands to operate on the cluster and the
databases was issued in a different generation of the cluster,
to ensure that commands that operate on the databases will not
survive across a cluster database recovery. After a recovery,
all old outstanding commands will automatically become invalid.
Sometimes this number will be shown as "INVALID". This only
means that the ctdbd daemon has started but it has not yet
merged with the cluster through a recovery. All nodes start with
generation "INVALID" and are not assigned a real generation id
until they have successfully been merged with a cluster through
a recovery.
VNNMAP
The list of Virtual Node Numbers. This is a list of all nodes
that actively participates in the cluster and that share the
workload of hosting the Clustered TDB database records. Only
nodes that are participating in the vnnmap can become lmaster or
dmaster for a database record.
Recovery mode
This is the current recovery mode of the cluster. There are two
possible modes:
NORMAL - The cluster is fully operational.
RECOVERY - The cluster databases have all been frozen, pausing
all services while the cluster awaits a recovery process to
complete. A recovery process should finish within seconds. If a
cluster is stuck in the RECOVERY state this would indicate a
cluster malfunction which needs to be investigated.
Once the recovery master detects an inconsistency, for example a
node becomes disconnected/connected, the recovery daemon will
trigger a cluster recovery process, where all databases are
remerged across the cluster. When this process starts, the
recovery master will first "freeze" all databases to prevent
applications such as samba from accessing the databases and it
will also mark the recovery mode as RECOVERY.
When CTDBD starts up, it will start in RECOVERY mode. Once the
node has been merged into a cluster and all databases have been
recovered, the node mode will change into NORMAL mode and the
databases will be "thawed", allowing samba to access the
databases again.
Recovery master
This is the cluster node that is currently designated as the
recovery master. This node is responsible of monitoring the
consistency of the cluster and to perform the actual recovery
process when reqired.
Only one node at a time can be the designated recovery master.
Which node is designated the recovery master is decided by an
election process in the recovery daemons running on each node.
Example: ctdb status
Example output:
Number of nodes:4
pnn:0 11.1.2.200 OK (THIS NODE)
pnn:1 11.1.2.201 OK
pnn:2 11.1.2.202 OK
pnn:3 11.1.2.203 OK
Generation:1362079228
Size:4
hash:0 lmaster:0
hash:1 lmaster:1
hash:2 lmaster:2
hash:3 lmaster:3
Recovery mode:NORMAL (0)
Recovery master:0
recmaster
This command shows the pnn of the node which is currently the
recmaster.
uptime
This command shows the uptime for the ctdb daemon. When the last
recovery or ip-failover completed and how long it took. If the
"duration" is shown as a negative number, this indicates that there is
a recovery/failover in progress and it started that many seconds ago.
Example: ctdb uptime
Example output:
Current time of node : Thu Oct 29 10:38:54 2009
Ctdbd start time : (000 16:54:28) Wed Oct 28 17:44:26 2009
Time of last recovery/failover: (000 16:53:31) Wed Oct 28 17:45:23 2009
Duration of last recovery/failover: 2.248552 seconds
listnodes
This command shows lists the ip addresses of all the nodes in the
cluster.
Example: ctdb listnodes
Example output:
10.0.0.71
10.0.0.72
10.0.0.73
10.0.0.74
ping
This command will "ping" all CTDB daemons in the cluster to verify that
they are processing commands correctly.
Example: ctdb ping
Example output:
response from 0 time=0.000054 sec (3 clients)
response from 1 time=0.000144 sec (2 clients)
response from 2 time=0.000105 sec (2 clients)
response from 3 time=0.000114 sec (2 clients)
ip
This command will display the list of public addresses that are
provided by the cluster and which physical node is currently serving
this ip. By default this command will ONLY show those public addresses
that are known to the node itself. To see the full list of all public
ips across the cluster you must use "ctdb ip -n all".
Example: ctdb ip
Example output:
Number of addresses:4
12.1.1.1 0
12.1.1.2 1
12.1.1.3 2
12.1.1.4 3
scriptstatus
This command displays which scripts where run in the previous
monitoring cycle and the result of each script. If a script failed with
an error, causing the node to become unhealthy, the output from that
script is also shown.
Example: ctdb scriptstatus
Example output:
7 scripts were executed last monitoring cycle
00.ctdb Status:OK Duration:0.056 Tue Mar 24 18:56:57 2009
10.interface Status:OK Duration:0.077 Tue Mar 24 18:56:57 2009
11.natgw Status:OK Duration:0.039 Tue Mar 24 18:56:57 2009
20.multipathd Status:OK Duration:0.038 Tue Mar 24 18:56:57 2009
31.clamd Status:DISABLED
40.vsftpd Status:OK Duration:0.045 Tue Mar 24 18:56:57 2009
41.httpd Status:OK Duration:0.039 Tue Mar 24 18:56:57 2009
50.samba Status:ERROR Duration:0.082 Tue Mar 24 18:56:57 2009
OUTPUT:ERROR: Samba tcp port 445 is not responding
disablescript <script>
This command is used to disable an eventscript.
This will take effect the next time the eventscripts are being executed
so it can take a short while until this is reflected in ´scriptstatus´.
enablescript <script>
This command is used to enable an eventscript.
This will take effect the next time the eventscripts are being executed
so it can take a short while until this is reflected in ´scriptstatus´.
getvar <name>
Get the runtime value of a tuneable variable.
Example: ctdb getvar MaxRedirectCount
Example output:
MaxRedirectCount = 3
setvar <name> <value>
Set the runtime value of a tuneable variable.
Example: ctdb setvar MaxRedirectCount 5
listvars
List all tuneable variables.
Example: ctdb listvars
Example output:
MaxRedirectCount = 5
SeqnumFrequency = 1
ControlTimeout = 60
TraverseTimeout = 20
KeepaliveInterval = 2
KeepaliveLimit = 3
MaxLACount = 7
RecoverTimeout = 5
RecoverInterval = 1
ElectionTimeout = 3
TakeoverTimeout = 5
MonitorInterval = 15
EventScriptTimeout = 20
RecoveryGracePeriod = 60
RecoveryBanPeriod = 300
lvsmaster
This command shows which node is currently the LVSMASTER. The LVSMASTER
is the node in the cluster which drives the LVS system and which
receives all incoming traffic from clients.
LVS is the mode where the entire CTDB/Samba cluster uses a single ip
address for the entire cluster. In this mode all clients connect to one
specific node which will then multiplex/loadbalance the clients evenly
onto the other nodes in the cluster. This is an alternative to using
public ip addresses. See the manpage for ctdbd for more information
about LVS.
lvs
This command shows which nodes in the cluster are currently active in
the LVS configuration. I.e. which nodes we are currently loadbalancing
the single ip address across.
LVS will by default only loadbalance across those nodes that are both
LVS capable and also HEALTHY. Except if all nodes are UNHEALTHY in
which case LVS will loadbalance across all UNHEALTHY nodes as well. LVS
will never use nodes that are DISCONNECTED, STOPPED, BANNED or
DISABLED.
Example output:
2:10.0.0.13
3:10.0.0.14
getcapabilities
This command shows the capabilities of the current node. Please see
manpage for ctdbd for a full list of all capabilities and more detailed
description.
RECMASTER and LMASTER capabilities are primarily used when CTDBD is
used to create a cluster spanning across WAN links. In which case ctdbd
acts as a WAN accelerator.
LVS capabile means that the node is participating in LVS, a mode where
the entire CTDB cluster uses one single ip address for the entire
cluster instead of using public ip address failover. This is an
alternative to using a loadbalancing layer-4 switch.
Example output:
RECMASTER: YES
LMASTER: YES
LVS: NO
statistics
Collect statistics from the CTDB daemon about how many calls it has
served.
Example: ctdb statistics
Example output:
CTDB version 1
num_clients 3
frozen 0
recovering 0
client_packets_sent 360489
client_packets_recv 360466
node_packets_sent 480931
node_packets_recv 240120
keepalive_packets_sent 4
keepalive_packets_recv 3
node
req_call 2
reply_call 2
req_dmaster 0
reply_dmaster 0
reply_error 0
req_message 42
req_control 120408
reply_control 360439
client
req_call 2
req_message 24
req_control 360440
timeouts
call 0
control 0
traverse 0
total_calls 2
pending_calls 0
lockwait_calls 0
pending_lockwait_calls 0
memory_used 5040
max_hop_count 0
max_call_latency 4.948321 sec
max_lockwait_latency 0.000000 sec
statisticsreset
This command is used to clear all statistics counters in a node.
Example: ctdb statisticsreset
getreclock
This command is used to show the filename of the reclock file that is
used.
Example output:
Reclock file:/gpfs/.ctdb/shared
setreclock [filename]
This command is used to modify, or clear, the file that is used as the
reclock file at runtime. When this command is used, the reclock file
checks are disabled. To re-enable the checks the administrator needs to
activate the "VerifyRecoveryLock" tunable using "ctdb setvar".
If run with no parameter this will remove the reclock file completely.
If run with a parameter the parameter specifies the new filename to use
for the recovery lock.
This command only affects the runtime settings of a ctdb node and will
be lost when ctdb is restarted. For persistent changes to the reclock
file setting you must edit /etc/sysconfig/ctdb.
getdebug
Get the current debug level for the node. the debug level controls what
information is written to the log file.
The debug levels are mapped to the corresponding syslog levels. When a
debug level is set, only those messages at that level and higher levels
will be printed.
The list of debug levels from highest to lowest are :
EMERG ALERT CRIT ERR WARNING NOTICE INFO DEBUG
setdebug <debuglevel>
Set the debug level of a node. This controls what information will be
logged.
The debuglevel is one of EMERG ALERT CRIT ERR WARNING NOTICE INFO DEBUG
getpid
This command will return the process id of the ctdb daemon.
disable
This command is used to administratively disable a node in the cluster.
A disabled node will still participate in the cluster and host
clustered TDB records but its public ip address has been taken over by
a different node and it no longer hosts any services.
enable
Re-enable a node that has been administratively disabled.
stop
This command is used to administratively STOP a node in the cluster. A
STOPPED node is connected to the cluster but will not host any public
ip addresse, nor does it participate in the VNNMAP. The difference
between a DISABLED node and a STOPPED node is that a STOPPED node does
not host any parts of the database which means that a recovery is
required to stop/continue nodes.
continue
Re-start a node that has been administratively stopped.
addip <public_ip/mask> <iface>
This command is used to add a new public ip to a node during runtime.
This allows public addresses to be added to a cluster without having to
restart the ctdb daemons.
Note that this only updates the runtime instance of ctdb. Any changes
will be lost next time ctdb is restarted and the public addresses file
is re-read. If you want this change to be permanent you must also
update the public addresses file manually.
delip <public_ip>
This command is used to remove a public ip from a node during runtime.
If this public ip is currently hosted by the node it being removed
from, the ip will first be failed over to another node, if possible,
before it is removed.
Note that this only updates the runtime instance of ctdb. Any changes
will be lost next time ctdb is restarted and the public addresses file
is re-read. If you want this change to be permanent you must also
update the public addresses file manually.
moveip <public_ip> <node>
This command can be used to manually fail a public ip address to a
specific node.
In order to manually override the "automatic" distribution of public ip
addresses that ctdb normally provides, this command only works when you
have changed the tunables for the daemon to:
DeterministicIPs = 0
NoIPFailback = 1
shutdown
This command will shutdown a specific CTDB daemon.
recover
This command will trigger the recovery daemon to do a cluster recovery.
ipreallocate
This command will force the recovery master to perform a full ip
reallocation process and redistribute all ip addresses. This is useful
to "reset" the allocations back to its default state if they have been
changed using the "moveip" command. While a "recover" will also perform
this reallocation, a recovery is much more hevyweight since it will
also rebuild all the databases.
setlmasterrole <on|off>
This command is used ot enable/disable the LMASTER capability for a
node at runtime. This capability determines whether or not a node can
be used as an LMASTER for records in the database. A node that does not
have the LMASTER capability will not show up in the vnnmap.
Nodes will by default have this capability, but it can be stripped off
nodes by the setting in the sysconfig file or by using this command.
Once this setting has been enabled/disabled, you need to perform a
recovery for it to take effect.
See also "ctdb getcapabilities"
setrecmasterrole <on|off>
This command is used ot enable/disable the RECMASTER capability for a
node at runtime. This capability determines whether or not a node can
be used as an RECMASTER for the cluster. A node that does not have the
RECMASTER capability can not win a recmaster election. A node that
already is the recmaster for the cluster when the capability is
stripped off the node will remain the recmaster until the next cluster
election.
Nodes will by default have this capability, but it can be stripped off
nodes by the setting in the sysconfig file or by using this command.
See also "ctdb getcapabilities"
killtcp <srcip:port> <dstip:port>
This command will kill the specified TCP connection by issuing a TCP
RST to the srcip:port endpoint. This is a command used by the ctdb
eventscripts.
gratiousarp <ip> <interface>
This command will send out a gratious arp for the specified interface
through the specified interface. This command is mainly used by the
ctdb eventscripts.
reloadnodes
This command is used when adding new nodes, or removing existing nodes
from an existing cluster.
Procedure to add a node:
1, To expand an existing cluster, first ensure with ´ctdb status´ that
all nodes are up and running and that they are all healthy. Do not try
to expand a cluster unless it is completely healthy!
2, On all nodes, edit /etc/ctdb/nodes and add the new node as the last
entry to the file. The new node MUST be added to the end of this file!
3, Verify that all the nodes have identical /etc/ctdb/nodes files after
you edited them and added the new node!
4, Run ´ctdb reloadnodes´ to force all nodes to reload the nodesfile.
5, Use ´ctdb status´ on all nodes and verify that they now show the
additional node.
6, Install and configure the new node and bring it online.
Procedure to remove a node:
1, To remove a node from an existing cluster, first ensure with ´ctdb
status´ that all nodes, except the node to be deleted, are up and
running and that they are all healthy. Do not try to remove a node from
a cluster unless the cluster is completely healthy!
2, Shutdown and poerwoff the node to be removed.
3, On all other nodes, edit the /etc/ctdb/nodes file and comment out
the node to be removed. Do not delete the line for that node, just
comment it out by adding a ´#´ at the beginning of the line.
4, Run ´ctdb reloadnodes´ to force all nodes to reload the nodesfile.
5, Use ´ctdb status´ on all nodes and verify that the deleted node no
longer shows up in the list..
tickle <srcip:port> <dstip:port>
This command will will send a TCP tickle to the source host for the
specified TCP connection. A TCP tickle is a TCP ACK packet with an
invalid sequence and acknowledge number and will when received by the
source host result in it sending an immediate correct ACK back to the
other end.
TCP tickles are useful to "tickle" clients after a IP failover has
occured since this will make the client immediately recognize the TCP
connection has been disrupted and that the client will need to
reestablish. This greatly speeds up the time it takes for a client to
detect and reestablish after an IP failover in the ctdb cluster.
gettickles <ip>
This command is used to show which TCP connections are registered with
CTDB to be "tickled" if there is a failover.
repack [max_freelist]
Over time, when records are created and deleted in a TDB, the TDB list
of free space will become fragmented. This can lead to a slowdown in
accessing TDB records. This command is used to defragment a TDB
database and pruning the freelist.
If [max_freelist] is specified, then a database will only be repacked
if it has more than this number of entries in the freelist.
During repacking of the database, the entire TDB database will be
locked to prevent writes. If samba tries to write to a record in the
database during a repack operation, samba will block until the
repacking has completed.
This command can be disruptive and can cause samba to block for the
duration of the repack operation. In general, a repack operation will
take less than one second to complete.
A repack operation will only defragment the local TDB copy of the CTDB
database. You need to run this command on all of the nodes to repack a
CTDB database completely.
Example: ctdb repack 1000
By default, this operation is issued from the 00.ctdb event script
every 5 minutes.
vacuum [max_records]
Over time CTDB databases will fill up with empty deleted records which
will lead to a progressive slow down of CTDB database access. This
command is used to prune all databases and delete all empty records
from the cluster.
By default, vacuum will delete all empty records from all databases. If
[max_records] is specified, the command will only delete the first
[max_records] empty records for each database.
Vacuum only deletes records where the local node is the lmaster. To
delete all records from the entire cluster you need to run a vacuum
from each node. This command is not disruptive. Samba is unaffected and
will still be able to read/write records normally while the database is
being vacuumed.
Example: ctdb vacuum
By default, this operation is issued from the 00.ctdb event script
every 5 minutes.
backupdb <database> <file>
This command can be used to copy the entire content of a database out
to a file. This file can later be read back into ctdb using the
restoredb command. This is mainly useful for backing up persistent
databases such as secrets.tdb and similar.
restoredb <file>
This command restores a persistent database that was previously backed
up using backupdb.
DEBUGGING COMMANDS
These commands are primarily used for CTDB development and testing and
should not be used for normal administration.
process-exists <pid>
This command checks if a specific process exists on the CTDB host. This
is mainly used by Samba to check if remote instances of samba are still
running or not.
getdbmap
This command lists all clustered TDB databases that the CTDB daemon has
attached to. Some databases are flagged as PERSISTENT, this means that
the database stores data persistently and the data will remain across
reboots. One example of such a database is secrets.tdb where
information about how the cluster was joined to the domain is stored.
Most databases are not persistent and only store the state information
that the currently running samba daemons need. These databases are
always wiped when ctdb/samba starts and when a node is rebooted.
Example: ctdb getdbmap
Example output:
Number of databases:10
dbid:0x435d3410 name:notify.tdb path:/var/ctdb/notify.tdb.0
dbid:0x42fe72c5 name:locking.tdb path:/var/ctdb/locking.tdb.0 dbid:0x1421fb78 name:brlock.tdb path:/var/ctdb/brlock.tdb.0
dbid:0x17055d90 name:connections.tdb path:/var/ctdb/connections.tdb.0
dbid:0xc0bdde6a name:sessionid.tdb path:/var/ctdb/sessionid.tdb.0
dbid:0x122224da name:test.tdb path:/var/ctdb/test.tdb.0
dbid:0x2672a57f name:idmap2.tdb path:/var/ctdb/persistent/idmap2.tdb.0 PERSISTENT
dbid:0xb775fff6 name:secrets.tdb path:/var/ctdb/persistent/secrets.tdb.0 PERSISTENT
dbid:0xe98e08b6 name:group_mapping.tdb path:/var/ctdb/persistent/group_mapping.tdb.0 PERSISTENT
dbid:0x7bbbd26c name:passdb.tdb path:/var/ctdb/persistent/passdb.tdb.0 PERSISTENT
catdb <dbname>
This command will dump a clustered TDB database to the screen. This is
a debugging command.
getmonmode
This command returns the monutoring mode of a node. The monitoring mode
is either ACTIVE or DISABLED. Normally a node will continously monitor
that all other nodes that are expected are in fact connected and that
they respond to commands.
ACTIVE - This is the normal mode. The node is actively monitoring all
other nodes, both that the transport is connected and also that the
node responds to commands. If a node becomes unavailable, it will be
marked as DISCONNECTED and a recovery is initiated to restore the
cluster.
DISABLED - This node is not monitoring that other nodes are available.
In this mode a node failure will not be detected and no recovery will
be performed. This mode is useful when for debugging purposes one wants
to attach GDB to a ctdb process but wants to prevent the rest of the
cluster from marking this node as DISCONNECTED and do a recovery.
setmonmode <0|1>
This command can be used to explicitely disable/enable monitoring mode
on a node. The main purpose is if one wants to attach GDB to a running
ctdb daemon but wants to prevent the other nodes from marking it as
DISCONNECTED and issuing a recovery. To do this, set monitoring mode to
0 on all nodes before attaching with GDB. Remember to set monitoring
mode back to 1 afterwards.
attach <dbname>
This is a debugging command. This command will make the CTDB daemon
create a new CTDB database and attach to it.
dumpmemory
This is a debugging command. This command will make the ctdb daemon to
write a fill memory allocation map to standard output.
rddumpmemory
This is a debugging command. This command will dump the talloc memory
allocation tree for the recovery daemon to standard output.
freeze
This command will lock all the local TDB databases causing clients that
are accessing these TDBs such as samba3 to block until the databases
are thawed.
This is primarily used by the recovery daemon to stop all samba daemons
from accessing any databases while the database is recovered and
rebuilt.
thaw
Thaw a previously frozen node.
eventscript <arguments>
This is a debugging command. This command can be used to manually
invoke and run the eventscritps with arbitrary arguments.
ban <bantime|0>
Administratively ban a node for bantime seconds. A bantime of 0 means
that the node should be permanently banned.
A banned node does not participate in the cluster and does not host any
records for the clustered TDB. Its ip address has been taken over by an
other node and no services are hosted.
Nodes are automatically banned if they are the cause of too many
cluster recoveries.
unban
This command is used to unban a node that has either been
administratively banned using the ban command or has been automatically
banned by the recovery daemon.
SEE ALSO
ctdbd(1), onnode(1) http://ctdb.samba.org/
COPYRIGHT/LICENSE
Copyright (C) Andrew Tridgell 2007
Copyright (C) Ronnie sahlberg 2007
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or (at
your option) any later version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, see http://www.gnu.org/licenses/.
10/29/2009