ctdb

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

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>: +

OPTIONS

-n <pnn>: This specifies the physical node number on which to execute the command. Default is to run the command on the daemon running on the local host. @@ -28,13 +28,13 @@ 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. @@ -57,7 +57,7 @@ in a cluster like a node that is ok. Some interfaces to serve public ip addresses are down, but at least one interface is up. See also "ctdb ifaces". -

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. @@ -78,10 +78,10 @@ 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. @@ -101,7 +101,7 @@ 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 @@ -123,9 +123,9 @@ 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 @@ -134,7 +134,7 @@ 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 @@ -143,7 +143,7 @@ Duration of last recovery/failover: 2.248552 seconds 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 @@ -154,7 +154,7 @@ 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) -

ifaces

This command will display the list of network interfaces, which could host public addresses, along with their status.

@@ -177,13 +177,13 @@ name:eth2 link:up references:1 :eth4:0:0 :eth3:1:1 :eth2:1:1 -

setifacelink <iface> <status>

This command will set the status of a network interface. The status needs to be "up" or "down". This is typically used in the 10.interfaces script in the "monitor" event.

Example: ctdb setifacelink eth0 up -

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 @@ -213,7 +213,7 @@ Public IPs on node 0 :172.31.92.83:0:eth5:eth5:eth4,eth5: :172.31.92.84:1::eth5:eth4,eth5: :172.31.92.85:0:eth5:eth5:eth4,eth5: -

ipinfo <ip>

This command will display details about the specified public addresses.

Example: ctdb ipinfo 172.31.92.85 @@ -226,7 +226,7 @@ CurrentNode:0 NumInterfaces:2 Interface[1]: Name:eth4 Link:down References:0 Interface[2]: Name:eth5 Link:up References:2 (active) -

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 @@ -243,15 +243,15 @@ Interface[2]: Name:eth5 Link:up References:2 (active) 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 @@ -259,11 +259,11 @@ Interface[2]: Name:eth5 Link:up References:2 (active) 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 @@ -312,7 +312,7 @@ VacuumMaxInterval = 600 MaxQueueDropMsg = 1000 UseStatusEvents = 0 AllowUnhealthyDBRead = 0 -

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. @@ -323,7 +323,7 @@ AllowUnhealthyDBRead = 0 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. @@ -338,7 +338,7 @@ AllowUnhealthyDBRead = 0

 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. @@ -357,7 +357,7 @@ AllowUnhealthyDBRead = 0 RECMASTER: YES LMASTER: YES LVS: NO -

statistics

Collect statistics from the CTDB daemon about how many calls it has served.

Example: ctdb statistics @@ -399,23 +399,23 @@ CTDB version 1 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. @@ -425,42 +425,42 @@ Reclock file:/gpfs/.ctdb/shared 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.

moveip <public_ip> <node>

This command can be used to manually fail a public ip address to a specific node.

@@ -471,14 +471,14 @@ Reclock file:/gpfs/.ctdb/shared 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. @@ -486,21 +486,21 @@ Reclock file:/gpfs/.ctdb/shared Once this setting has been enabled/disabled, you need to perform a recovery for it to take effect.

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.

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: @@ -526,7 +526,7 @@ Reclock file:/gpfs/.ctdb/shared 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. + 2, Shutdown and poweroff 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.

@@ -534,7 +534,7 @@ Reclock file:/gpfs/.ctdb/shared

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 @@ -546,10 +546,10 @@ Reclock file:/gpfs/.ctdb/shared 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.

@@ -564,7 +564,7 @@ Reclock file:/gpfs/.ctdb/shared 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.

@@ -580,17 +580,17 @@ Reclock file:/gpfs/.ctdb/shared Example: ctdb vacuum

By default, this operation is issued from the 00.ctdb event script every 5 minutes. -

backupdb <dbname> <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> [<dbname>]

This command restores a persistent database that was previously backed up using backupdb. By default the data will be restored back into the same database as it was created from. By specifying dbname you can restore the data into a different database. -

wipedb <dbname>

This command can be used to remove all content of a database. -

getlog <level>

In addition to the normal loggign to a log file, CTDBD also keeps a in-memory ringbuffer containing the most recent log entries for all log levels (except DEBUG). @@ -603,14 +603,14 @@ This is mainly useful for backing up persistent databases such as secrets.tdb an

This command extracts all messages of level or lower log level from memory and prints it to the screen. -

clearlog

This command clears the in-memory logging ringbuffer. -

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.

If a PERSISTENT database is not in a healthy state the database is @@ -650,7 +650,7 @@ dbid:0xb775fff6 name:secrets.tdb path:/var/ctdb/persistent/secrets.tdb.0 PERSIST

 :ID:Name:Path:Persistent:Unhealthy:
 :0x7bbbd26c:passdb.tdb:/var/ctdb/persistent/passdb.tdb.0:1:0:
-

getdbstatus <dbname>

This command displays more details about a database.

Example: ctdb getdbstatus test.tdb.0 @@ -672,47 +672,47 @@ name: registry.tdb path: /var/ctdb/persistent/registry.tdb.0 PERSISTENT: yes HEALTH: NO-HEALTHY-NODES - ERROR - Backup of corrupted TDB in '/var/ctdb/persistent/registry.tdb.0.corrupted.20091208091949.0Z' -

catdb <dbname>

This command will dump a clustered TDB database to the screen. This is a debugging command. -

dumpdbbackup <backup-file>

This command will dump the content of database backup to the screen (similar to ctdb catdb). 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 continuously 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 explicitly 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> [persistent]

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. -

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 another 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. -