CTDB(1)
💖 ctdb(1) — CTDB management utility 💖
okay bestie, welcome to ctdb from the CTDB suite — the clustered TDB database situation. this is the management utility page. grab a snack, we're going in 🍓
✨ SYNOPSIS ✨
ctdb [OPTION...] {COMMAND} [COMMAND-ARGS]🦋 DESCRIPTION 🦋
ctdb is the little utility that lets you view and manage a CTDB cluster. that's literally the whole vibe, but let's get into the vocabulary first because the rest of this page uses these words constantly and you will need them.
so when we talk about nodes in a cluster, here's the lingo:
- PNN — Physical Node Number. it's an integer that describes the node in the cluster. the first node has physical node number
0. in a cluster. 💅 (yes that trailing "in a cluster" is in the original, we keep it, periodt) - PNN-LIST — this is either a single PNN, a comma-separate list of PNNs, or the word
"all". that's it. that's the whole format. iykyk.
and when commands reference a database, here's that vocabulary:
- DB — this is either a database name, like
locking.tdb, or a database ID, like"0x42fe72c5". - DB-LIST — a space separated list of at least one DB.
👑 OPTIONS 👑
-n PNN— query the node specified by PNN for whatever info you're asking for. default behaviour is to query the daemon running on the local host, so if you don't say anything, that's who's answering.-Y— produce output in machine readable form so scripts can parse it without crying. uses a field delimiter of':'. ngl, not all commands support this option, so don't come for me if yours doesn't.-x SEPARATOR— use SEPARATOR to delimit fields in machine readable output. this implies-Y(it comes as a package deal 🎀).-X— also machine readable output for easy script parsing, but this one uses a field delimiter of'|'. again, not all commands support it.
this is equivalent to "-x|" and it avoids some shell quoting issues, which is the real reason you'd pick it.
-t TIMEOUT— tellsctdbto wait up to TIMEOUT seconds for a response to most commands sent to the CTDB daemon. default is10seconds.-T TIMELIMIT— TIMELIMIT is the maximum run time (in seconds) for thectdbcommand. if TIMELIMIT gets exceeded, thectdbcommand will terminate with an error. default is120seconds.-? --help— print some help text to the screen. classic.--usage— print usage information to the screen. slightly different from help, don't ask, just use it.-d --debug=DEBUGLEVEL— change the debug level for the command. default isNOTICE.
🌈 ADMINISTRATIVE COMMANDS 🌈
these are the commands you use to monitor and administer a CTDB cluster. let's go. 🫶
💜 pnn
this command displays the PNN of the current node. short, sweet, iconic.
💜 status
this one shows the current status of all CTDB nodes, based on information from the queried node.
Note: if the queried node is INACTIVE then the status might not be current. so like… don't trust it blindly, bestie.
here's what you actually get out of it:
Node status
this includes the number of physical nodes and the status of each node. see ctdb(7) for info about node states.
Generation
the generation id is a number that indicates the current generation of a cluster instance. every time a cluster goes through a reconfiguration or a recovery, its generation id gets changed.
this number doesn't have any particular meaning other than to keep track of when a cluster has gone through a recovery. it's a random number that represents the current instance of a ctdb cluster and its databases. the CTDB daemon uses this number internally so it can tell when commands to operate on the cluster and the databases were issued in a different generation of the cluster — that way commands that operate on the databases won't survive across a cluster database recovery. after a recovery, all old outstanding commands automatically become invalid. sorry to those commands. 💀
sometimes this number shows up as "INVALID". that only means the ctdbd daemon has started but it hasn't yet merged with the cluster through a recovery. all nodes start with generation "INVALID" and don't get a real generation id until they've successfully been merged with a cluster through a recovery.
Virtual Node Number (VNN) map
consists of the number of virtual nodes and the mapping from virtual node numbers to physical node numbers. only nodes that are participating in the VNN map can become lmaster for database records.
Recovery mode
this is the current recovery mode of the cluster. there are two possible modes:
NORMAL— the cluster is fully operational. slay.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 theRECOVERYstate that would indicate a cluster malfunction which needs to be investigated. 🚨
once the leader 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 leader 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 the CTDB daemon starts up, it starts in RECOVERY mode. once the node has been merged into a cluster and all databases have been recovered, the node mode changes into NORMAL mode and the databases get "thawed", letting samba access the databases again. character development.
Leader
this is the cluster node currently designated as the leader. this node is responsible of monitoring the consistency of the cluster and to perform the actual recovery process when required.
only one node at a time can be the designated leader. which node gets designated is decided by an election process in the recovery daemons running on each node. very democracy, very main character energy. 👑
Example
# ctdb status
Number of nodes:4
pnn:0 192.168.2.200 OK (THIS NODE)
pnn:1 192.168.2.201 OK
pnn:2 192.168.2.202 OK
pnn:3 192.168.2.203 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)
Leader:0💜 nodestatus [PNN-LIST]
this command is similar to the status command. it displays the "node status" subset of output. the main differences are:
- the exit code is the bitwise-OR of the flags for each specified node, while
ctdb statusexits with0if it was able to retrieve status for all nodes. ctdb statusprovides status information for all nodes.ctdb nodestatusdefaults to providing status for only the current node. if PNN-LIST is provided then status is given for the indicated node(s).
a common invocation in scripts is ctdb nodestatus all to check whether all nodes in a cluster are healthy. ✨
Example
# ctdb nodestatus
pnn:0 10.0.0.30 OK (THIS NODE)
# ctdb nodestatus all
Number of nodes:2
pnn:0 10.0.0.30 OK (THIS NODE)
pnn:1 10.0.0.31 OK💜 leader
this command shows the pnn of the node which is currently the leader.
Note: if the queried node is INACTIVE then the status might not be current. (yes, same warning as before, it really means it.)
💜 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
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. (yes, "shows lists", that's the original phrasing, we're not editing it, we respect the text 💅)
Example
# ctdb listnodes
192.168.2.200
192.168.2.201
192.168.2.202
192.168.2.203💜 natgw {leader|list|status}
this command shows different aspects of NAT gateway status. for an overview of CTDB's NAT gateway functionality, please see the NAT GATEWAY section in ctdb(7).
leader— show the PNN and private IP address of the current NAT gateway leader node.
example output:
1 192.168.2.201list— list the private IP addresses of nodes in the current NAT gateway group, annotating the leader node.
example output:
192.168.2.200
192.168.2.201 LEADER
192.168.2.202
192.168.2.203status— list the nodes in the current NAT gateway group and their status.
example output:
pnn:0 192.168.2.200 UNHEALTHY (THIS NODE)
pnn:1 192.168.2.201 OK
pnn:2 192.168.2.202 OK
pnn:3 192.168.2.203 OK💜 ping
this command will "ping" specified CTDB nodes in the cluster to verify that they are running. 🏓
Example
# ctdb ping
response from 0 time=0.000054 sec (3 clients)💜 ifaces
this command will display the list of network interfaces, which could host public addresses, along with their status.
Example
# ctdb ifaces
Interfaces on node 0
name:eth5 link:up references:2
name:eth4 link:down references:0
name:eth3 link:up references:1
name:eth2 link:up references:1
# ctdb -X ifaces
|Name|LinkStatus|References|
|eth5|1|2|
|eth4|0|0|
|eth3|1|1|
|eth2|1|1|💜 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 all".
Example
# ctdb ip -v
Public IPs on node 0
172.31.91.82 node[1] active[] available[eth2,eth3] configured[eth2,eth3]
172.31.91.83 node[0] active[eth3] available[eth2,eth3] configured[eth2,eth3]
172.31.91.84 node[1] active[] available[eth2,eth3] configured[eth2,eth3]
172.31.91.85 node[0] active[eth2] available[eth2,eth3] configured[eth2,eth3]
172.31.92.82 node[1] active[] available[eth5] configured[eth4,eth5]
172.31.92.83 node[0] active[eth5] available[eth5] configured[eth4,eth5]
172.31.92.84 node[1] active[] available[eth5] configured[eth4,eth5]
172.31.92.85 node[0] active[eth5] available[eth5] configured[eth4,eth5]okay that's the end of part 1, but bestie we are NOT done, there's more coming — don't go anywhere 💖
And lowkey here's the tea when you ask for -X and -v at the same time — it hands you a whole lil' dashboard:
# ctdb -X ip -v
|Public IP|Node|ActiveInterface|AvailableInterfaces|ConfiguredInterfaces|
|172.31.91.82|1||eth2,eth3|eth2,eth3|
|172.31.91.83|0|eth3|eth2,eth3|eth2,eth3|
|172.31.91.84|1||eth2,eth3|eth2,eth3|
|172.31.91.85|0|eth2|eth2,eth3|eth2,eth3|
|172.31.92.82|1||eth5|eth4,eth5|
|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> 💖
Okay so ipinfo is the vibe check for ONE specific public address. Hand it an IP and it spills every single detail it's got about that address, bestie.
Example
# ctdb ipinfo 172.31.92.85
Public IP[172.31.92.85] info on node 0
IP:172.31.92.85
CurrentNode:0
NumInterfaces:2
Interface[1]: Name:eth4 Link:down References:0
Interface[2]: Name:eth5 Link:up References:2 (active)event run|status|script list|script enable|script disable ✨
This whole command family is how you boss around the event daemon and peek at what all the various events are doing. Observe 👑
Heads up though: the commands below all need a component named, and in the current version the only valid component is legacy. Periodt.
run TIMEOUT COMPONENT EVENT [ARGUMENTS]
This one lets you manually kick off a specific EVENT inside COMPONENT, optionally with ARGUMENTS. The event gets a max of TIMEOUT seconds to do its thing. And if TIMEOUT is 0? No time limit at all, go off, run forever, no cap.
status COMPONENT EVENT
Shows you the last execution status of the given EVENT in COMPONENT.
The command bails out with an exit status that matches the overall status of the event it just showed you — so the exit code is literally the event's whole mood.
The output is the list of event scripts that got run. Every single line gives you the name, status, duration and start time for each script. And whatever the script printed? Yeah, that gets shown too. Fr fr.
Example #1
# ctdb event status legacy monitor
00.ctdb OK 0.014 Sat Apr 15 19:39:11 2024
01.reclock OK 0.013 Sat Apr 15 19:39:11 2024
05.system OK 0.029 Sat Apr 15 19:39:11 2024
10.interface OK 0.037 Sat Apr 15 19:39:11 2024
11.natgw OK 0.011 Sat Apr 15 19:39:11 2024
11.routing OK 0.007 Sat Apr 15 19:39:11 2024
13.per_ip_routing OK 0.007 Sat Apr 15 19:39:11 2024
20.multipathd OK 0.007 Sat Apr 15 19:39:11 2024
31.clamd OK 0.007 Sat Apr 15 19:39:11 2024
40.vsftpd OK 0.013 Sat Apr 15 19:39:11 2024
41.httpd OK 0.018 Sat Apr 15 19:39:11 2024
49.winbind OK 0.023 Sat Apr 15 19:39:11 2024
50.samba OK 0.100 Sat Apr 15 19:39:12 2024
60.nfs OK 0.376 Sat Apr 15 19:39:12 2024
70.iscsi OK 0.009 Sat Apr 15 19:39:12 2024
91.lvs OK 0.007 Sat Apr 15 19:39:12 2024
95.database OK 0.009 Sat Apr 15 19:39:12 2024Example #2
# ctdb event status legacy monitor
00.ctdb OK 0.011 Sat Apr 15 19:40:46 2024
01.reclock OK 0.010 Sat Apr 15 19:40:46 2024
05.system OK 0.030 Sat Apr 15 19:40:46 2024
10.interface OK 0.041 Sat Apr 15 19:40:46 2024
11.natgw OK 0.008 Sat Apr 15 19:40:46 2024
11.routing OK 0.007 Sat Apr 15 19:40:46 2024
13.per_ip_routing OK 0.007 Sat Apr 15 19:40:46 2024
20.multipathd OK 0.007 Sat Apr 15 19:40:46 2024
31.clamd OK 0.007 Sat Apr 15 19:40:46 2024
40.vsftpd OK 0.013 Sat Apr 15 19:40:46 2024
41.httpd OK 0.015 Sat Apr 15 19:40:46 2024
49.winbind OK 0.022 Sat Apr 15 19:40:46 2024
50.samba ERROR 0.077 Sat Apr 15 19:40:46 2024
OUTPUT: ERROR: samba tcp port 445 is not respondingscript list COMPONENT
Lists every event script that's available in COMPONENT. The ones that are enabled get a lil' * flag next to them, slay.
So generally, event scripts come from CTDB itself. BUT — local or 3rd party event scripts can also exist, and those show up in their own separate section after the CTDB-provided ones. Grouped and cute, we love to see it.
Example
# ctdb event script list legacy
* 00.ctdb
* 01.reclock
* 05.system
* 10.interface
11.natgw
11.routing
13.per_ip_routing
20.multipathd
31.clamd
40.vsftpd
41.httpd
* 49.winbind
* 50.samba
* 60.nfs
70.iscsi
91.lvs
* 95.database
* 02.localscript enable COMPONENT SCRIPT
Enables the named event SCRIPT in COMPONENT. Lowkey important: only enabled scripts get executed when any event runs. So if it's not enabled, it's just sitting there, not doing anything, like a phone on Do Not Disturb.
script disable COMPONENT SCRIPT
Disables the named event SCRIPT in COMPONENT. This stops that script from executing whenever any event runs. Bye bestie 💅
scriptstatus 🦋
This is just an alias for ctdb event status legacy [EVENT], where EVENT defaults to monitor.
⚠️ Real talk: this command is deprecated. It only exists for backward compatibility. Use ctdb event status instead, okay? Do it for me 🙏
listvars 👑
Lists all the tuneable variables — with one tiny exception: the values of obsolete tunables like VacuumMinInterval do NOT show up here. Those ancient lil' guys can only be fetched explicitly, using the ctdb getvar command. iykyk.
Example
# ctdb listvars
SeqnumInterval = 1000
ControlTimeout = 60
TraverseTimeout = 20
KeepaliveInterval = 5
KeepaliveLimit = 5
RecoverTimeout = 120
RecoverInterval = 1
ElectionTimeout = 3
TakeoverTimeout = 9
MonitorInterval = 15
TickleUpdateInterval = 20
EventScriptTimeout = 30
MonitorTimeoutCount = 20
RecoveryGracePeriod = 120
RecoveryBanPeriod = 300
DatabaseHashSize = 100001
DatabaseMaxDead = 5
RerecoveryTimeout = 10
EnableBans = 1
NoIPFailback = 0
VerboseMemoryNames = 0
RecdPingTimeout = 60
RecdFailCount = 10
LogLatencyMs = 0
RecLockLatencyMs = 1000
RecoveryDropAllIPs = 120
VacuumInterval = 10
VacuumMaxRunTime = 120
RepackLimit = 10000
VacuumFastPathCount = 60
MaxQueueDropMsg = 1000000
AllowUnhealthyDBRead = 0
StatHistoryInterval = 1
DeferredAttachTO = 120
AllowClientDBAttach = 1
RecoverPDBBySeqNum = 1
DeferredRebalanceOnNodeAdd = 300
FetchCollapse = 1
HopcountMakeSticky = 50
StickyDuration = 600
StickyPindown = 200
NoIPTakeover = 0
DBRecordCountWarn = 100000
DBRecordSizeWarn = 10000000
DBSizeWarn = 100000000
PullDBPreallocation = 10485760
LockProcessesPerDB = 200
RecBufferSizeLimit = 1000000
QueueBufferSize = 1024
IPAllocAlgorithm = 2getvar <NAME> 🔍
Grabs the runtime value of a tuneable variable. One variable, one number, no drama.
Example
# ctdb getvar MonitorInterval
MonitorInterval = 15setvar <NAME> <VALUE> ✨
Sets the runtime value of a tuneable variable. Live changes only, bestie — go change a number and watch the world react.
Example
# ctdb setvar MonitorInterval 20lvs {leader|list|status} 🧚
This one shows you different angles of LVS status. And if you want the full LVS lore? Go read the LVS section in ctdb(7). Iconic crossover.
leader
Shows the PNN of whichever node is currently the LVS leader.
Example output:
2list
Lists the LVS nodes that are currently usable.
Example output:
2 10.0.0.13
3 10.0.0.14status
Lists the nodes in the current LVS group, plus their status. A little roster, if you will.
Example output:
pnn:0 10.0.0.11 UNHEALTHY (THIS NODE)
pnn:1 10.0.0.12 UNHEALTHY
pnn:2 10.0.0.13 OK
pnn:3 10.0.0.14 OKgetcapabilities 🌈
Shows the capabilities of the node you're currently on. Want the full explainer? The CAPABILITIES section in ctdb(7) has you covered.
Example output:
LEADER: YES
LMASTER: YESstatistics 💀
Collects statistics straight from the CTDB daemon about how many calls it has served. And if you're like "wait what does each field even mean" — all that info lives in ctdb-statistics(7). Go look, it's giving reference material.
Example ✨
# ctdb statistics
CTDB version 1
Current time of statistics : Tue Mar 8 15:18:51 2016
Statistics collected since : (003 21:31:32) Fri Mar 4 17:47:19 2016
num_clients 9
frozen 0
recovering 0
num_recoveries 2
client_packets_sent 8170534
client_packets_recv 7166132
node_packets_sent 16549998
node_packets_recv 5244418
keepalive_packets_sent 201969
keepalive_packets_recv 201969
node
req_call 26
reply_call 0
req_dmaster 9
reply_dmaster 12
reply_error 0
req_message 1339231
req_control 8177506
reply_control 6831284
client
req_call 15
req_message 334809
req_control 6831308
timeouts
call 0
control 0
traverse 0
locks
num_calls 8
num_current 0
num_pending 0
num_failed 0
total_calls 15
pending_calls 0
childwrite_calls 0
pending_childwrite_calls 0
memory_used 394879
max_hop_count 1
total_ro_delegations 0
total_ro_revokes 0
hop_count_buckets: 8 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0
lock_buckets: 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0
locks_latency MIN/AVG/MAX 0.010005/0.010418/0.011010 sec out of 8
reclock_ctdbd MIN/AVG/MAX 0.002538/0.002538/0.002538 sec out of 1
reclock_recd MIN/AVG/MAX 0.000000/0.000000/0.000000 sec out of 0
call_latency MIN/AVG/MAX 0.000044/0.002142/0.011702 sec out of 15
childwrite_latency MIN/AVG/MAX 0.000000/0.000000/0.000000 sec out of 0🧹 statisticsreset
Bestie. This one wipes ALL the statistics counters on a node. Every. Single. One. Gone. Clean slate energy ✨
Example:
ctdb statisticsreset📊 dbstatistics DB
Shows you the stats for one specific database, DB. Wanna know what all those fields even mean? Go read ctdb-statistics(7), it's all laid out there 🫶
Example:
# ctdb dbstatistics locking.tdb
DB Statistics: locking.tdb
ro_delegations 0
ro_revokes 0
locks
total 14356
failed 0
current 0
pending 0
hop_count_buckets: 28087 2 1 0 0 0 0 0 0 0 0 0 0 0 0 0
lock_buckets: 0 14188 38 76 32 19 3 0 0 0 0 0 0 0 0 0
locks_latency MIN/AVG/MAX 0.001066/0.012686/4.202292 sec out of 14356
vacuum_latency MIN/AVG/MAX 0.000472/0.002207/15.243570 sec out of 224530
Num Hot Keys: 1
Count:8 Key:ff5bd7cb3ee3822edc1f0000000000000000000000000000🔐 getreclock
Shows you the deets on the recovery lock — if there even is one. No lock? Nothing to show 💀
Example output:
/clusterfs/.ctdb/recovery.lock🔍 getdebug
Grabs the current debug level for the node. Real talk: the debug level is literally what decides how much info gets written into the log file.
The debug levels map straight onto the matching syslog levels. Once a debug level is set, only messages at that level and above get printed — so nothing below it makes the cut 🚫
The list of debug levels, from highest to lowest, is:
ERROR WARNING NOTICE INFO DEBUG
🎚️ setdebug DEBUGLEVEL
Sets the debug level for a node, which is what controls what actually gets logged.
The debuglevel is one of ERROR WARNING NOTICE INFO DEBUG
🆔 getpid
Hands you back the process id of the ctdb daemon. That's it, that's the whole vibe ✨
🚫 disable
Administratively disables a node in the cluster. A disabled node? Still in the cluster, still hosting clustered TDB records — but its public ip address has been taken over by a different node, and it no longer hosts any services. She's there, she's just not doing the most 💅
✅ enable
Re-enables a node that was administratively disabled. Welcome back bb 🫶
🛑 stop
Administratively STOPs a node in the cluster. A STOPPED node is still connected to the cluster, but it will not host any public ip addresses, and it doesn't participate in the VNNMAP either.
Here's the difference between a DISABLED node and a STOPPED node, and it's a big one: a STOPPED node hosts zero parts of the database — which means a recovery is required to stop/continue nodes. No cap.
▶️ continue
Re-starts a node that was administratively stopped. She's back online, periodt.
➕ addip IPADDR/mask IFACE
Adds a brand new public ip to a node at runtime. You should follow it up with a ctdb ipreallocate. This is the whole point: it lets you add public addresses to a cluster without having to restart the ctdb daemons 🎀
Heads up bestie: this only updates the runtime instance ofctdb. Any changes will be lost next timectdbis restarted and the public addresses file is re-read. Want it to stick? You also gotta update the public addresses file manually, fr fr.
➖ delip IPADDR
Flags IPADDR for deletion from a node at runtime. Should be followed by a ctdb ipreallocate. If IPADDR is currently hosted by the node it's being removed from, this makes sure the IP first gets failed over to another node, if possible, and then that it's actually removed. Clean exit strategy 👑
Same warning, same energy: 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. Want it permanent? Update the public addresses file manually too.
🔀 moveip IPADDR PNN
Manually fails a public ip address over to a specific node. Iconic.
But here's the catch — in order to manually override the "automatic" distribution of public ip addresses that ctdb normally provides, this command only works when you've changed the tunables for the daemon to:
IPAllocAlgorithm != 0
NoIPFailback = 1💤 shutdown
Shuts down a specific CTDB daemon. Night night 😴
👑 setlmasterrole on|off
Turns the LMASTER capability for a node on or off at runtime. This capability is what determines whether or not a node can be used as an LMASTER for records in the database. A node without the LMASTER capability will not show up in the vnnmap.
Nodes have this capability by default, but it can get stripped off nodes by the setting in the sysconfig file, or by using this command.
Once you've enabled/disabled this setting, you need to perform a recovery for it to actually take effect.
See also ctdb getcapabilities
🌟 setleaderrole on|off
Turns the LEADER capability for a node on or off at runtime. This capability determines whether or not a node can be elected leader of the cluster. A node without the LEADER capability cannot be elected leader. And if the current leader has this capability removed? Bestie, an election will occur 💀
Nodes have this capability enabled by default, but it can be removed via the cluster:leader capability configuration setting, or by using this command.
See also ctdb getcapabilities
🔄 reloadnodes
This is the one you reach for when you're adding new nodes to an existing cluster, or removing existing nodes from one.
Procedure to add nodes:
- To expand an existing cluster, first ensure with
ctdb statusthat all nodes are up and running and that they are all healthy. Do not try to expand a cluster unless it is completely healthy! - On all nodes, edit
/etc/ctdb/nodesand add the new nodes at the end of this file. - Verify that all the nodes have identical
/etc/ctdb/nodesfiles after adding the new nodes. - Run
ctdb reloadnodesto force all nodes to reload the nodes file. - Use
ctdb statuson all nodes and verify that they now show the additional nodes. - Install and configure the new node and bring it online.
Procedure to remove nodes:
- To remove nodes from an existing cluster, first ensure with
ctdb statusthat all nodes, except the node to be deleted, are up and running and that they are all healthy. Do not try to remove nodes from a cluster unless the cluster is completely healthy! - Shutdown and power off the node to be removed.
- On all other nodes, edit the
/etc/ctdb/nodesfile and comment out the nodes to be removed. Do not delete the lines for the deleted nodes, just comment them out by adding a#at the beginning of the lines. - Run
ctdb reloadnodesto force all nodes to reload the nodes file. - Use
ctdb statuson all nodes and verify that the deleted nodes are no longer listed.
📡 reloadips [PNN-LIST]
Reloads the public addresses configuration file on the nodes you specify. When it finishes, addresses will be reconfigured and reassigned across the cluster as necessary. Slay.
Lowkey annoying limitation: this command is currently unable to make changes to the netmask or interfaces associated with existing addresses. Those kinds of changes have to be made in 2 steps — delete the addresses in question and re-add them. And unfortunately, that will disrupt connections to the changed addresses. Sorry bestie 😭
🗺️ getdbmap
Lists off every single clustered TDB database that the CTDB daemon has attached to.
Some databases get flagged as PERSISTENT — that means the database stores data persistently and the data will remain across reboots. One example of such a database is secrets.tdb, which is where the info about how the cluster was joined to the domain lives.
Some databases get flagged as REPLICATED — that means the data in that database is replicated across all the nodes. But the data will not remain across reboots. This type of database is what CTDB uses to store its own internal state.
If a PERSISTENT database is not in a healthy state, the database gets flagged as UNHEALTHY. If there's at least one completely healthy node running in the cluster, it's possible that the content gets restored automatically by a recovery run. Otherwise, an administrator needs to analyze the problem. No auto-fix for you, rip.
See also ctdb getdbstatus, ctdb backupdb, ctdb restoredb, ctdb dumpbackup, ctdb wipedb, ctdb setvar AllowUnhealthyDBRead 1 and (if samba or tdb-utils are installed) tdbtool check.
Most databases are lowkey not persistent, bestie — they only hold the state info that the samba daemons currently running need. These databases get wiped every single time ctdb/samba starts and whenever a node reboots. 💅 No cap.
Example:
# ctdb getdbmap
Number of databases:10
dbid:0x435d3410 name:notify.tdb path:/var/lib/ctdb/notify.tdb.0
dbid:0x42fe72c5 name:locking.tdb path:/var/lib/ctdb/locking.tdb.0
dbid:0x1421fb78 name:brlock.tdb path:/var/lib/ctdb/brlock.tdb.0
dbid:0x17055d90 name:connections.tdb path:/var/lib/ctdb/connections.tdb.0
dbid:0xc0bdde6a name:sessionid.tdb path:/var/lib/ctdb/sessionid.tdb.0
dbid:0x122224da name:test.tdb path:/var/lib/ctdb/test.tdb.0
dbid:0x2672a57f name:idmap2.tdb path:/var/lib/ctdb/persistent/idmap2.tdb.0 PERSISTENT
dbid:0xb775fff6 name:secrets.tdb path:/var/lib/ctdb/persistent/secrets.tdb.0 PERSISTENT
dbid:0xe98e08b6 name:group_mapping.tdb path:/var/lib/ctdb/persistent/group_mapping.tdb.0 PERSISTENT
dbid:0x7bbbd26c name:passdb.tdb path:/var/lib/ctdb/persistent/passdb.tdb.0 PERSISTENT
# ctdb getdbmap # example for unhealthy database
Number of databases:1
dbid:0xb775fff6 name:secrets.tdb path:/var/lib/ctdb/persistent/secrets.tdb.0 PERSISTENT UNHEALTHY
# ctdb -X getdbmap
|ID|Name|Path|Persistent|Unhealthy|
|0x7bbbd26c|passdb.tdb|/var/lib/ctdb/persistent/passdb.tdb.0|1|0|backupdb DB FILE
Copy the contents of database DB to FILE. You can read FILE back later using restoredb. This is mainly useful for backing up persistent databases such as secrets.tdb and similar. Basically a glow-up for your backups. ✨
restoredb FILE [DB]
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. Main character energy: you choose the destination. 💖
setdbreadonly DB
This command will enable the read-only record support for a database. This is an experimental feature to improve performance for contended records primarily in locking.tdb and brlock.tdb. When enabling this feature you must set it on all nodes in the cluster. Periodt. 👑
setdbsticky DB
This command will enable the sticky record support for the specified database. This is an experimental feature to improve performance for contended records primarily in locking.tdb and brlock.tdb. When enabling this feature you must set it on all nodes in the cluster. Don't skip a node, bestie. 🫶
💖 INTERNAL COMMANDS 💖
Internal commands are used by CTDB's scripts and are not required for managing a CTDB cluster. Their parameters and behaviour are subject to change. So like, they're backstage, not the main show. 🙈
gettickles IPADDR
Show TCP connections that are registered with CTDB to be "tickled" if there is a failover. Tickles are a whole vibe. ✨
gratarp IPADDR INTERFACE
Send out a gratuitous ARP for the specified interface through the specified interface. This command is mainly used by the ctdb eventscripts. Yes, interface through interface. That's what it says. 😭
pdelete DB KEY
Delete KEY from DB. Bye bestie. 💅
pfetch DB KEY
Print the value associated with KEY in DB. It's giving "what does the key say?" 🔑
pstore DB KEY FILE
Store KEY in DB with contents of FILE as the associated value. Slay. 💾
ptrans DB [FILE]
Read a list of key-value pairs, one per line from FILE, and store them in DB using a single transaction. An empty value is equivalent to deleting the given key.
The key and value should be separated by spaces or tabs. Each key/value should be a printable string enclosed in double-quotes. So "key" "value" energy. ✨
runstate [setup|first_recovery|startup|running]
Print the runstate of the specified node. Runstates are used to serialise important state transitions in CTDB, particularly during startup.
If one or more optional runstate arguments are specified then the node must be in one of these runstates for the command to succeed. Vibe check: pass or fail. 💅
Example:
# ctdb runstate
RUNNINGsetifacelink IFACE up|down
Set the internal state of network interface IFACE. This is typically used in the 10.interface script in the "monitor" event.
Example: ctdb setifacelink eth0 up. Get that interface in line. 🔥
tickle
Read a list of TCP connections, one per line, from standard input and send a TCP tickle to the source host for each connection. A connection is specified as:
SRC-IPADDR:SRC-PORT DST-IPADDR:DST-PORTA single connection can be specified on the command-line rather than on standard input.
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 occurred 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. So it's basically poking the client so it wakes up. 👉👈
version
Display the CTDB version. Iconic. 💅
😈 DEBUGGING COMMANDS 😈
These commands are primarily used for CTDB development and testing and should not be used for normal administration. So don't use them in prod unless you're literally a CTDB dev. Periodt. 🚫
✨ OPTIONS ✨
--print-emptyrecords— This enables printing of empty records when dumping databases with thecatdb,cattdbanddumpdbbackupcommands. Records with empty data segment are considered deleted by ctdb and cleaned by the vacuuming mechanism, so this switch can come in handy for debugging the vacuuming behaviour. Lowkey useful for vacuum vibes. 🧹--print-datasize— This lets database dumps (catdb,cattdb,dumpdbbackup) print the size of the record data instead of dumping the data contents. So you get the size, not the tea. 📏--print-lmaster— This letscatdbprint the lmaster for each record. Lmaster said 👑.--print-hash— This lets database dumps (catdb,cattdb,dumpdbbackup) print the hash for each record. Hash it out. #️⃣--print-recordflags— This letscatdbanddumpdbbackupprint the record flags for each record. Note thatcattdbalways prints the flags. So cattdb is extra. 💅
process-exists PID [SRVID]
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. When the optional SRVID argument is specified, the command checks if a specific process exists on the CTDB host and has registered for specified SRVID. It's giving "are you still there?" 👀
getdbstatus DB
This command displays more details about a database.
Example:
# ctdb getdbstatus test.tdb.0
dbid: 0x122224da
name: test.tdb
path: /var/lib/ctdb/test.tdb.0
PERSISTENT: no
HEALTH: OK
# ctdb getdbstatus registry.tdb # with a corrupted TDB
dbid: 0xf2a58948
name: registry.tdb
path: /var/lib/ctdb/persistent/registry.tdb.0
PERSISTENT: yes
HEALTH: NO-HEALTHY-NODES - ERROR - Backup of corrupted TDB in '/var/lib/ctdb/persistent/registry.tdb.0.corrupted.20091208091949.0Z'catdb DB
Print a dump of the clustered TDB database DB. It's like cat but for a clustered TDB. Meow. 🐱
cattdb DB
Print a dump of the contents of the local TDB database DB. Local tea only. ☕
dumpdbbackup FILE
Print a dump of the contents from database backup FILE, similar to catdb. Backup dump, no cap. 💾
wipedb DB
Remove all contents of database DB. Yeet. 🗑️
recover
This command will trigger the recovery daemon to do a cluster recovery. Recovery era. 💪
ipreallocate, sync
This command will force the leader 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 heavyweight since it will also rebuild all the databases.
attach DBNAME [persistent|replicated]
Create a new CTDB database called DBNAME and attach to it on all nodes. New DB just dropped. 💅
detach DB-LIST
Detach specified non-persistent database(s) from the cluster. This command will disconnect specified database(s) on all nodes in the cluster. This command should only be used when none of the specified database(s) are in use.
All nodes should be active and tunable AllowClientDBAccess should be disabled on all nodes before detaching databases. So don't detach while people are using it, bestie. 🚫
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. Talloc tree, slay. 🌳
ban BANTIME
Administratively ban a node for BANTIME seconds. The node will be unbanned after BANTIME seconds have elapsed.
A banned node does not participate in the cluster. It does not host any records for the clustered TDB and does not host any public IP addresses.
Nodes are automatically banned if they misbehave. For example, a node may be banned if it causes too many cluster recoveries.
To administratively exclude a node from a cluster use the stop command. So ban is temporary, stop is forever. 💀
unban
This command is used to unban a node that has either been administratively banned using the ban command or has been automatically banned. Welcome back, bestie. 🫶
👀 SEE ALSO 👀
✍️ AUTHOR ✍️
This documentation was written by Ronnie Sahlberg, Amitay Isaacs, Martin Schwenke. Legends. 💖
©️ COPYRIGHT ©️
Copyright © 2007 Andrew Tridgell, Ronnie Sahlberg
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.
This page is from CTDB(1), dated 09/13/2026 — source/version: ctdb. Bye bestie, go slay some clusters. 💅✨
📦 where this page comes from
- ubuntu stonking: ctdb 2:4.24.7+dfsg-1ubuntu1 · GPL-3.0-or-later
- 💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms