1 <?xml version="1.0" encoding="iso-8859-1"?>
2 <!DOCTYPE refentry PUBLIC "-//Samba-Team//DTD DocBook V4.2-Based Variant V1.0//EN" "http://www.samba.org/samba/DTD/samba-doc">
3 <refentry id="ctdbd.1">
6 <refentrytitle>ctdbd</refentrytitle>
7 <manvolnum>1</manvolnum>
8 <refmiscinfo class="source">ctdb</refmiscinfo>
9 <refmiscinfo class="manual">CTDB - clustered TDB database</refmiscinfo>
14 <refname>ctdbd</refname>
15 <refpurpose>The CTDB cluster daemon</refpurpose>
20 <command>ctdbd</command>
24 <command>ctdbd</command>
25 <arg choice="opt">-? --help</arg>
26 <arg choice="opt">-d --debug=<INTEGER></arg>
27 <arg choice="req">--dbdir=<directory></arg>
28 <arg choice="req">--dbdir-persistent=<directory></arg>
29 <arg choice="opt">--event-script-dir=<directory></arg>
30 <arg choice="opt">-i --interactive</arg>
31 <arg choice="opt">--listen=<address></arg>
32 <arg choice="opt">--logfile=<filename></arg>
33 <arg choice="opt">--lvs</arg>
34 <arg choice="req">--nlist=<filename></arg>
35 <arg choice="opt">--no-lmaster</arg>
36 <arg choice="opt">--no-recmaster</arg>
37 <arg choice="opt">--nosetsched</arg>
38 <arg choice="req">--notification-script=<filename></arg>
39 <arg choice="opt">--public-addresses=<filename></arg>
40 <arg choice="opt">--public-interface=<interface></arg>
41 <arg choice="req">--reclock=<filename></arg>
42 <arg choice="opt">--single-public-ip=<address></arg>
43 <arg choice="opt">--socket=<filename></arg>
44 <arg choice="opt">--start-as-disabled</arg>
45 <arg choice="opt">--start-as-stopped</arg>
46 <arg choice="opt">--syslog</arg>
47 <arg choice="opt">--log-ringbuf-size=<num-entries></arg>
48 <arg choice="opt">--torture</arg>
49 <arg choice="opt">--transport=<STRING></arg>
50 <arg choice="opt">--usage</arg>
55 <refsect1><title>DESCRIPTION</title>
57 ctdbd is the main ctdb daemon.
60 ctdbd provides a clustered version of the TDB database with automatic rebuild/recovery of the databases upon nodefailures.
63 Combined with a cluster filesystem ctdbd provides a full HA environment for services such as clustered Samba and NFS as well as other services.
66 ctdbd provides monitoring of all nodes in the cluster and automatically reconfigures the cluster and recovers upon node failures.
69 ctdbd is the main component in clustered Samba that provides a high-availability load-sharing CIFS server cluster.
75 <title>OPTIONS</title>
78 <varlistentry><term>-? --help</term>
81 Print some help text to the screen.
86 <varlistentry><term>-d --debug=<DEBUGLEVEL></term>
89 This option sets the debuglevel on the ctdbd daemon which controls what will be written to the logfile. The default is 0 which will only log important events and errors. A larger number will provide additional logging.
94 <varlistentry><term>--dbdir=<directory></term>
97 This is the directory on local storage where ctdbd keeps the local
98 copy of the TDB databases. This directory is local for each node and should not be stored on the shared cluster filesystem.
101 This directory would usually be /var/ctdb .
106 <varlistentry><term>--dbdir-persistent=<directory></term>
109 This is the directory on local storage where ctdbd keeps the local
110 copy of the persistent TDB databases. This directory is local for each node and should not be stored on the shared cluster filesystem.
113 This directory would usually be /etc/ctdb/persistent .
118 <varlistentry><term>--event-script-dir=<directory></term>
121 This option is used to specify the directory where the CTDB event
125 This will normally be /etc/ctdb/events.d which is part of the ctdb distribution.
130 <varlistentry><term>-i --interactive</term>
133 By default ctdbd will detach itself from the shell and run in
134 the background as a daemon. This option makes ctdbd to start in interactive mode.
139 <varlistentry><term>--listen=<address></term>
142 This specifies which ip address ctdb will bind to. By default ctdbd will bind to the first address it finds in the /etc/ctdb/nodes file and which is also present on the local system in which case you do not need to provide this option.
145 This option is only required when you want to run multiple ctdbd daemons/nodes on the same physical host in which case there would be multiple entries in /etc/ctdb/nodes what would match a local interface.
150 <varlistentry><term>--logfile=<filename></term>
153 This is the file where ctdbd will write its log. This is usually /var/log/log.ctdb .
158 <varlistentry><term>--lvs</term>
161 This option is used to activate the LVS capability on a CTDB node.
162 Please see the LVS section.
167 <varlistentry><term>--nlist=<filename></term>
170 This file contains a list of the private ip addresses of every node in the cluster. There is one line/ip address for each node. This file must be the same for all nodes in the cluster.
173 This file is usually /etc/ctdb/nodes .
178 <varlistentry><term>--no-lmaster</term>
181 This argument specifies that this node can NOT become an lmaster
182 for records in the database. This means that it will never show up
183 in the vnnmap. This feature is primarily used for making a cluster
184 span across a WAN link and use CTDB as a WAN-accelerator.
187 Please see the "remote cluster nodes" section for more information.
192 <varlistentry><term>--no-recmaster</term>
195 This argument specifies that this node can NOT become a recmaster
196 for the database. This feature is primarily used for making a cluster
197 span across a WAN link and use CTDB as a WAN-accelerator.
200 Please see the "remote cluster nodes" section for more information.
205 <varlistentry><term>--nosetsched</term>
208 This is a ctdbd debugging option. this option is only used when
212 Normally ctdb will change its scheduler to run as a real-time
213 process. This is the default mode for a normal ctdbd operation
214 to gurarantee that ctdbd always gets the cpu cycles that it needs.
217 This option is used to tell ctdbd to NOT run as a real-time process
218 and instead run ctdbd as a normal userspace process.
219 This is useful for debugging and when you want to run ctdbd under
220 valgrind or gdb. (You don't want to attach valgrind or gdb to a
226 <varlistentry><term>--notification-script=<filename></term>
229 This specifies a script which will be invoked by ctdb when certain
230 state changes occur in ctdbd and when you may want to trigger this
231 to run certain scripts.
234 This file is usually /etc/ctdb/notify.sh .
237 See the NOTIFICATION SCRIPT section below for more information.
242 <varlistentry><term>--public_addresses=<filename></term>
245 When used with IP takeover this specifies a file containing the public ip addresses to use on the cluster. This file contains a list of ip addresses netmasks and interfaces. When ctdb is operational it will distribute these public ip addresses evenly across the available nodes.
248 This is usually the file /etc/ctdb/public_addresses
253 <varlistentry><term>--public-interface=<interface></term>
256 This option tells ctdb which interface to attach public-addresses
257 to and also where to attach the single-public-ip when used.
260 This is only required when using public ip addresses and only when
261 you don't specify the interface explicitly in /etc/ctdb/public_addresses or when you are using --single-public-ip.
264 If you omit this argument when using public addresses or single public ip, ctdb will not be able to send out Gratious ARPs correctly or be able to kill tcp connections correctly which will lead to application failures.
269 <varlistentry><term>--reclock=<filename></term>
272 This is the name of the lock file stored of the shared cluster filesystem that ctdbd uses to prevent split brains from occuring.
273 This file must be stored on shared storage.
276 It is possible to run CTDB without a reclock file, but then there
277 will be no protection against split brain if the network becomes
278 partitioned. Using CTDB without a reclock file is strongly
284 <varlistentry><term>--socket=<filename></term>
287 This specifies the name of the domain socket that ctdbd will create. This socket is used for local clients to attach to and communicate with the ctdbd daemon.
290 The default is /tmp/ctdb.socket . You only need to use this option if you plan to run multiple ctdbd daemons on the same physical host.
295 <varlistentry><term>--start-as-disabled</term>
298 This makes the ctdb daemon to be DISABLED when it starts up.
301 As it is DISABLED it will not get any of the public ip addresses
302 allocated to it, and thus this allow you to start ctdb on a node
303 without causing any ip address to failover from other nodes onto
307 When used, the administrator must keep track of when nodes start and
308 manually enable them again using the "ctdb enable" command, or else
309 the node will not host any services.
312 A node that is DISABLED will not host any services and will not be
313 reachable/used by any clients.
318 <varlistentry><term>--start-as-stopped</term>
321 This makes the ctdb daemon to be STOPPED when it starts up.
324 A node that is STOPPED does not host any public addresses. It is not part of the VNNMAP so it does act as an LMASTER. It also has all databases locked in recovery mode until restarted.
327 To restart and activate a STOPPED node, the command "ctdb continue" is used.
330 A node that is STOPPED will not host any services and will not be
331 reachable/used by any clients.
336 <varlistentry><term>--syslog</term>
339 Send all log messages to syslog instead of to the ctdb logfile.
344 <varlistentry><term>--log-ringbuf-size=<num-entries></term>
347 In addition to the normal loggign to a log file,
348 CTDBD also keeps a in-memory ringbuffer containing the most recent
349 log entries for all log levels (except DEBUG).
352 This is useful since it allows for keeping continuous logs to a file
353 at a reasonable non-verbose level, but shortly after an incident has
354 occured, a much more detailed log can be pulled from memory. This
355 can allow you to avoid having to reproduce an issue due to the
356 on-disk logs being of insufficient detail.
359 This in-memory ringbuffer contains a fixed number of the most recent
360 entries. This is settable at startup either through the
361 --log-ringbuf-size argument, or preferably by using
362 CTDB_LOG_RINGBUF_SIZE in the sysconfig file.
365 Use the "ctdb getlog" command to access this log.
371 <varlistentry><term>--torture</term>
374 This option is only used for development and testing of ctdbd. It adds artificial errors and failures to the common codepaths in ctdbd to verify that ctdbd can recover correctly for failures.
377 You do NOT want to use this option unless you are developing and testing new functionality in ctdbd.
382 <varlistentry><term>--transport=<STRING></term>
385 This option specifies which transport to use for ctdbd internode communications. The default is "tcp".
388 Currently only "tcp" is supported but "infiniband" might be
389 implemented in the future.
394 <varlistentry><term>--usage</term>
397 Print useage information to the screen.
406 <refsect1><title>Private vs Public addresses</title>
408 When used for ip takeover in a HA environment, each node in a ctdb
409 cluster has multiple ip addresses assigned to it. One private and one or more public.
412 <refsect2><title>Private address</title>
414 This is the physical ip address of the node which is configured in
415 linux and attached to a physical interface. This address uniquely
416 identifies a physical node in the cluster and is the ip addresses
417 that ctdbd will use to communicate with the ctdbd daemons on the
418 other nodes in the cluster.
421 The private addresses are configured in /etc/ctdb/nodes
422 (unless the --nlist option is used) and contain one line for each
423 node in the cluster. Each line contains the private ip address for one
424 node in the cluster. This file must be the same on all nodes in the
428 Since the private addresses are only available to the network when the
429 corresponding node is up and running you should not use these addresses
430 for clients to connect to services provided by the cluster. Instead
431 client applications should only attach to the public addresses since
432 these are guaranteed to always be available.
435 When using ip takeover, it is strongly recommended that the private
436 addresses are configured on a private network physically separated
437 from the rest of the network and that this private network is dedicated
440 Example /etc/ctdb/nodes for a four node cluster:
441 <screen format="linespecific">
448 <refsect2><title>Public address</title>
450 A public address on the other hand is not attached to an interface.
451 This address is managed by ctdbd itself and is attached/detached to
452 a physical node at runtime.
455 The ctdb cluster will assign/reassign these public addresses across the
456 available healthy nodes in the cluster. When one node fails, its public address
457 will be migrated to and taken over by a different node in the cluster
458 to ensure that all public addresses are always available to clients as
459 long as there are still nodes available capable of hosting this address.
462 These addresses are not physically attached to a specific node.
463 The 'ctdb ip' command can be used to view the current assignment of
464 public addresses and which physical node is currently serving it.
467 On each node this file contains a list of the public addresses that
468 this node is capable of hosting.
469 The list also contain the netmask and the
470 interface where this address should be attached for the case where you
471 may want to serve data out through multiple different interfaces.
473 Example /etc/ctdb/public_addresses for a node that can host 4 public addresses:
474 <screen format="linespecific">
482 In most cases this file would be the same on all nodes in a cluster but
483 there are exceptions when one may want to use different files
486 Example: 4 nodes partitioned into two subgroups :
487 <screen format="linespecific">
488 Node 0:/etc/ctdb/public_addresses
492 Node 1:/etc/ctdb/public_addresses
496 Node 2:/etc/ctdb/public_addresses
500 Node 3:/etc/ctdb/public_addresses
505 In this example nodes 0 and 1 host two public addresses on the
506 10.1.1.x network while nodes 2 and 3 host two public addresses for the
510 Ip address 10.1.1.1 can be hosted by either of nodes 0 or 1 and will be
511 available to clients as long as at least one of these two nodes are
513 If both nodes 0 and node 1 become unavailable 10.1.1.1 also becomes
514 unavailable. 10.1.1.1 can not be failed over to node 2 or node 3 since
515 these nodes do not have this ip address listed in their public
522 <refsect1><title>Node status</title>
524 The current status of each node in the cluster can be viewed by the
525 'ctdb status' command.
528 There are five possible states for a node.
532 OK - This node is fully functional.
536 DISCONNECTED - This node could not be connected through the network
537 and is currently not particpating in the cluster. If there is a
538 public IP address associated with this node it should have been taken
539 over by a different node. No services are running on this node.
543 DISABLED - This node has been administratively disabled. This node is
544 still functional and participates in the CTDB cluster but its IP
545 addresses have been taken over by a different node and no services are
546 currently being hosted.
550 UNHEALTHY - A service provided by this node is malfunctioning and should
551 be investigated. The CTDB daemon itself is operational and participates
552 in the cluster. Its public IP address has been taken over by a different
553 node and no services are currently being hosted. All unhealthy nodes
554 should be investigated and require an administrative action to rectify.
558 BANNED - This node failed too many recovery attempts and has been banned
559 from participating in the cluster for a period of RecoveryBanPeriod
560 seconds. Any public IP address has been taken over by other nodes. This
561 node does not provide any services. All banned nodes should be
562 investigated and require an administrative action to rectify. This node
563 does not perticipate in the CTDB cluster but can still be communicated
564 with. I.e. ctdb commands can be sent to it.
568 STOPPED - A node that is stopped does not host any public ip addresses,
569 nor is it part of the VNNMAP. A stopped node can not become LVSMASTER,
571 This node does not perticipate in the CTDB cluster but can still be
572 communicated with. I.e. ctdb commands can be sent to it.
577 <title>PUBLIC TUNABLES</title>
579 These are the public tuneables that can be used to control how ctdb behaves.
582 <refsect2><title>MaxRedirectCount</title>
583 <para>Default: 3</para>
585 If we are not the DMASTER and need to fetch a record across the network
586 we first send the request to the LMASTER after which the record
587 is passed onto the current DMASTER. If the DMASTER changes before
588 the request has reached that node, the request will be passed onto the
589 "next" DMASTER. For very hot records that migrate rapidly across the
590 cluster this can cause a request to "chase" the record for many hops
591 before it catches up with the record.
593 this is how many hops we allow trying to chase the DMASTER before we
594 switch back to the LMASTER again to ask for new directions.
597 When chasing a record, this is how many hops we will chase the record
598 for before going back to the LMASTER to ask for new guidance.
602 <refsect2><title>SeqnumInterval</title>
603 <para>Default: 1000</para>
605 Some databases have seqnum tracking enabled, so that samba will be able
606 to detect asynchronously when there has been updates to the database.
607 Everytime a database is updated its sequence number is increased.
610 This tunable is used to specify in 'ms' how frequently ctdb will
611 send out updates to remote nodes to inform them that the sequence
616 <refsect2><title>ControlTimeout</title>
617 <para>Default: 60</para>
620 setting for timeout for when sending a control message to either the
621 local or a remote ctdb daemon.
625 <refsect2><title>TraverseTimeout</title>
626 <para>Default: 20</para>
628 This setting controls how long we allow a traverse process to run.
629 After this timeout triggers, the main ctdb daemon will abort the
630 traverse if it has not yet finished.
634 <refsect2><title>KeepaliveInterval</title>
635 <para>Default: 5</para>
637 How often in seconds should the nodes send keepalives to eachother.
641 <refsect2><title>KeepaliveLimit</title>
642 <para>Default: 5</para>
644 After how many keepalive intervals without any traffic should a node
645 wait until marking the peer as DISCONNECTED.
648 If a node has hung, it can thus take KeepaliveInterval*(KeepaliveLimit+1)
649 seconds before we determine that the node is DISCONNECTED and that we
650 require a recovery. This limitshould not be set too high since we want
651 a hung node to be detectec, and expunged from the cluster well before
652 common CIFS timeouts (45-90 seconds) kick in.
656 <refsect2><title>RecoverTimeout</title>
657 <para>Default: 20</para>
659 This is the default setting for timeouts for controls when sent from the
660 recovery daemon. We allow longer control timeouts from the recovery daemon
661 than from normal use since the recovery dameon often use controls that
662 can take a lot longer than normal controls.
666 <refsect2><title>RecoverInterval</title>
667 <para>Default: 1</para>
669 How frequently in seconds should the recovery daemon perform the
670 consistency checks that determine if we need to perform a recovery or not.
674 <refsect2><title>ElectionTimeout</title>
675 <para>Default: 3</para>
677 When electing a new recovery master, this is how many seconds we allow
678 the election to take before we either deem the election finished
679 or we fail the election and start a new one.
683 <refsect2><title>TakeoverTimeout</title>
684 <para>Default: 9</para>
686 This is how many seconds we allow controls to take for IP failover events.
690 <refsect2><title>MonitorInterval</title>
691 <para>Default: 15</para>
693 How often should ctdb run the event scripts to check for a nodes health.
697 <refsect2><title>TickleUpdateInterval</title>
698 <para>Default: 20</para>
700 How often will ctdb record and store the "tickle" information used to
701 kickstart stalled tcp connections after a recovery.
705 <refsect2><title>EventScriptTimeout</title>
706 <para>Default: 20</para>
708 How long should ctdb let an event script run before aborting it and
709 marking the node unhealthy.
713 <refsect2><title>EventScriptTimeoutCount</title>
714 <para>Default: 1</para>
716 How many events in a row needs to timeout before we flag the node UNHEALTHY.
717 This setting is useful if your scripts can not be written so that they
718 do not hang for benign reasons.
722 <refsect2><title>EventScriptUnhealthyOnTimeout</title>
723 <para>Default: 0</para>
725 This setting can be be used to make ctdb never become UNHEALTHY if your
726 eventscripts keep hanging/timing out.
730 <refsect2><title>RecoveryGracePeriod</title>
731 <para>Default: 120</para>
733 During recoveries, if a node has not caused recovery failures during the
734 last grace period, any records of transgressions that the node has caused
735 recovery failures will be forgiven. This resets the ban-counter back to
740 <refsect2><title>RecoveryBanPeriod</title>
741 <para>Default: 300</para>
743 If a node becomes banned causing repetitive recovery failures. The node will
744 eventually become banned from the cluster.
745 This controls how long the culprit node will be banned from the cluster
746 before it is allowed to try to join the cluster again.
747 Don't set to small. A node gets banned for a reason and it is usually due
748 to real problems with the node.
752 <refsect2><title>DatabaseHashSize</title>
753 <para>Default: 100001</para>
755 Size of the hash chains for the local store of the tdbs that ctdb manages.
759 <refsect2><title>DatabaseMaxDead</title>
760 <para>Default: 5</para>
762 How many dead records per hashchain in the TDB database do we allow before
763 the freelist needs to be processed.
767 <refsect2><title>RerecoveryTimeout</title>
768 <para>Default: 10</para>
770 Once a recovery has completed, no additional recoveries are permitted
771 until this timeout has expired.
775 <refsect2><title>EnableBans</title>
776 <para>Default: 1</para>
778 When set to 0, this disables BANNING completely in the cluster and thus
779 nodes can not get banned, even it they break. Don't set to 0 unless you
780 know what you are doing.
784 <refsect2><title>DeterministicIPs</title>
785 <para>Default: 0</para>
787 When enabled, this tunable makes ctdb try to keep public IP addresses
788 locked to specific nodes as far as possible. This makes it easier for
789 debugging since you can know that as long as all nodes are healthy
790 public IP X will always be hosted by node Y.
793 The cost of using deterministic IP address assignment is that it
794 disables part of the logic where ctdb tries to reduce the number of
795 public IP assignment changes in the cluster. This tunable may increase
796 the number of IP failover/failbacks that are performed on the cluster
801 <refsect2><title>LCP2PublicIPs</title>
802 <para>Default: 1</para>
804 When enabled this switches ctdb to use the LCP2 ip allocation
809 <refsect2><title>ReclockPingPeriod</title>
810 <para>Default: x</para>
816 <refsect2><title>NoIPFailback</title>
817 <para>Default: 0</para>
819 When set to 1, ctdb will not perform failback of IP addresses when a node
820 becomes healthy. Ctdb WILL perform failover of public IP addresses when a
821 node becomes UNHEALTHY, but when the node becomes HEALTHY again, ctdb
822 will not fail the addresses back.
825 Use with caution! Normally when a node becomes available to the cluster
826 ctdb will try to reassign public IP addresses onto the new node as a way
827 to distribute the workload evenly across the clusternode. Ctdb tries to
828 make sure that all running nodes have approximately the same number of
829 public addresses it hosts.
832 When you enable this tunable, CTDB will no longer attempt to rebalance
833 the cluster by failing IP addresses back to the new nodes. An unbalanced
834 cluster will therefore remain unbalanced until there is manual
835 intervention from the administrator. When this parameter is set, you can
836 manually fail public IP addresses over to the new node(s) using the
837 'ctdb moveip' command.
841 <refsect2><title>DisableIPFailover</title>
842 <para>Default: 0</para>
844 When enabled, ctdb will not perform failover or failback. Even if a
845 node fails while holding public IPs, ctdb will not recover the IPs or
846 assign them to another node.
849 When you enable this tunable, CTDB will no longer attempt to recover
850 the cluster by failing IP addresses over to other nodes. This leads to
851 a service outage until the administrator has manually performed failover
852 to replacement nodes using the 'ctdb moveip' command.
856 <refsect2><title>NoIPTakeover</title>
857 <para>Default: 0</para>
859 When set to 1, ctdb will allow ip addresses to be failed over onto this
860 node. Any ip addresses that the node currently hosts will remain on the
861 node but no new ip addresses can be failed over onto the node.
865 <refsect2><title>VerboseMemoryNames</title>
866 <para>Default: 0</para>
868 This feature consumes additional memory. when used the talloc library
869 will create more verbose names for all talloc allocated objects.
873 <refsect2><title>RecdPingTimeout</title>
874 <para>Default: 60</para>
876 If the main dameon has not heard a "ping" from the recovery dameon for
877 this many seconds, the main dameon will log a message that the recovery
878 daemon is potentially hung.
882 <refsect2><title>RecdFailCount</title>
883 <para>Default: 10</para>
885 If the recovery daemon has failed to ping the main dameon for this many
886 consecutive intervals, the main daemon will consider the recovery daemon
887 as hung and will try to restart it to recover.
891 <refsect2><title>LogLatencyMs</title>
892 <para>Default: 0</para>
894 When set to non-zero, this will make the main daemon log any operation that
895 took longer than this value, in 'ms', to complete.
896 These include "how long time a lockwait child process needed",
897 "how long time to write to a persistent database" but also
898 "how long did it take to get a response to a CALL from a remote node".
902 <refsect2><title>RecLockLatencyMs</title>
903 <para>Default: 1000</para>
905 When using a reclock file for split brain prevention, if set to non-zero
906 this tunable will make the recovery dameon log a message if the fcntl()
907 call to lock/testlock the recovery file takes longer than this number of
912 <refsect2><title>RecoveryDropAllIPs</title>
913 <para>Default: 120</para>
915 If we have been stuck in recovery, or stopped, or banned, mode for
916 this many seconds we will force drop all held public addresses.
920 <refsect2><title>verifyRecoveryLock</title>
921 <para>Default: 1</para>
923 Should we take a fcntl() lock on the reclock file to verify that we are the
924 sole recovery master node on the cluster or not.
928 <refsect2><title>DeferredAttachTO</title>
929 <para>Default: 120</para>
931 When databases are frozen we do not allow clients to attach to the
932 databases. Instead of returning an error immediately to the application
933 the attach request from the client is deferred until the database
934 becomes available again at which stage we respond to the client.
937 This timeout controls how long we will defer the request from the client
938 before timing it out and returning an error to the client.
942 <refsect2><title>HopcountMakeSticky</title>
943 <para>Default: 50</para>
945 If the database is set to 'STICKY' mode, using the 'ctdb setdbsticky'
946 command, any record that is seen as very hot and migrating so fast that
947 hopcount surpasses 50 is set to become a STICKY record for StickyDuration
948 seconds. This means that after each migration the record will be kept on
949 the node and prevented from being migrated off the node.
952 This setting allows to try to identify such records and stop them from
953 migrating across the cluster so fast. This will improve performance for
954 certain workloads, such as locking.tdb if many clients are opening/closing
955 the same file concurrently.
959 <refsect2><title>StickyDuration</title>
960 <para>Default: 600</para>
962 Once a record has been found to be fetch-lock hot and has been flagged to
963 become STICKY, this is for how long, in seconds, the record will be
964 flagged as a STICKY record.
968 <refsect2><title>StickyPindown</title>
969 <para>Default: 200</para>
971 Once a STICKY record has been migrated onto a node, it will be pinned down
972 on that node for this number of ms. Any request from other nodes to migrate
973 the record off the node will be deferred until the pindown timer expires.
977 <refsect2><title>MaxLACount</title>
978 <para>Default: 20</para>
980 When record content is fetched from a remote node, if it is only for
981 reading the record, pass back the content of the record but do not yet
982 migrate the record. Once MaxLACount identical requests from the
983 same remote node have been seen will the record be forcefully migrated
984 onto the requesting node. This reduces the amount of migration for a
985 database read-mostly workload at the expense of more frequent network
990 <refsect2><title>StatHistoryInterval</title>
991 <para>Default: 1</para>
993 Granularity of the statistics collected in the statistics history.
997 <refsect2><title>AllowClientDBAttach</title>
998 <para>Default: 1</para>
1000 When set to 0, clients are not allowed to attach to any databases.
1001 This can be used to temporarily block any new processes from attaching
1002 to and accessing the databases.
1006 <refsect2><title>RecoverPDBBySeqNum</title>
1007 <para>Default: 0</para>
1009 When set to non-zero, this will change how the recovery process for
1010 persistent databases ar performed. By default, when performing a database
1011 recovery, for normal as for persistent databases, recovery is
1012 record-by-record and recovery process simply collects the most recent
1013 version of every individual record.
1016 When set to non-zero, persistent databases will instead be recovered as
1017 a whole db and not by individual records. The node that contains the
1018 highest value stored in the record "__db_sequence_number__" is selected
1019 and the copy of that nodes database is used as the recovered database.
1023 <refsect2><title>FetchCollapse</title>
1024 <para>Default: 1</para>
1026 When many clients across many nodes try to access the same record at the
1027 same time this can lead to a fetch storm where the record becomes very
1028 active and bounces between nodes very fast. This leads to high CPU
1029 utilization of the ctdbd daemon, trying to bounce that record around
1030 very fast, and poor performance.
1033 This parameter is used to activate a fetch-collapse. A fetch-collapse
1034 is when we track which records we have requests in flight so that we only
1035 keep one request in flight from a certain node, even if multiple smbd
1036 processes are attemtping to fetch the record at the same time. This
1037 can improve performance and reduce CPU utilization for certain
1041 This timeout controls if we should collapse multiple fetch operations
1042 of the same record into a single request and defer all duplicates or not.
1048 <refsect1><title>LVS</title>
1050 LVS is a mode where CTDB presents one single IP address for the entire
1051 cluster. This is an alternative to using public IP addresses and round-robin
1052 DNS to loadbalance clients across the cluster.
1056 This is similar to using a layer-4 loadbalancing switch but with some restrictions.
1060 In this mode the cluster select a set of nodes in the cluster and loadbalance
1061 all client access to the LVS address across this set of nodes. This set of nodes are all LVS capable nodes that are HEALTHY, or if no HEALTHY nodes exists
1062 all LVS capable nodes regardless of health status.
1063 LVS will however never loadbalance traffic to nodes that are BANNED,
1064 STOPPED, DISABLED or DISCONNECTED. The "ctdb lvs" command is used to show
1065 which nodes are currently load-balanced across.
1069 One of the these nodes are elected as the LVSMASTER. This node receives all
1070 traffic from clients coming in to the LVS address and multiplexes it
1071 across the internal network to one of the nodes that LVS is using.
1072 When responding to the client, that node will send the data back
1073 directly to the client, bypassing the LVSMASTER node.
1074 The command "ctdb lvsmaster" will show which node is the current
1079 The path used for a client i/o is thus :
1080 <screen format="linespecific">
1081 (1) Client sends request packet to LVSMASTER
1082 (2) LVSMASTER passes the request on to one node across the internal network.
1083 (3) Selected node processes the request.
1084 (4) Node responds back to client.
1089 This means that all incoming traffic to the cluster will pass through
1090 one physical node, which limits scalability. You can send more data to the
1091 LVS address that one physical node can multiplex. This means that you
1092 should not use LVS if your I/O pattern is write-intensive since you will be
1093 limited in the available network bandwidth that node can handle.
1094 LVS does work wery well for read-intensive workloads where only smallish
1095 READ requests are going through the LVSMASTER bottleneck and the majority
1096 of the traffic volume (the data in the read replies) goes straight from
1097 the processing node back to the clients. For read-intensive i/o patterns you can acheive very high throughput rates in this mode.
1101 Note: you can use LVS and public addresses at the same time.
1104 <refsect2><title>Configuration</title>
1106 To activate LVS on a CTDB node you must specify CTDB_PUBLIC_INTERFACE and
1107 CTDB_LVS_PUBLIC_ADDRESS in /etc/sysconfig/ctdb.
1111 You must also specify the "--lvs" command line argument to ctdbd to activate LVS as a capability of the node. This should be done automatically for you by the /etc/init.d/ctdb script.
1116 <screen format="linespecific">
1117 CTDB_PUBLIC_INTERFACE=eth0
1118 CTDB_LVS_PUBLIC_IP=10.0.0.237
1125 If you use LVS, you must still have a real/permanent address configured
1126 for the public interface on each node. This address must be routable
1127 and the cluster nodes must be configured so that all traffic back to client
1128 hosts are routed through this interface. This is also required in order
1129 to allow samba/winbind on the node to talk to the domain controller.
1130 (we can not use the lvs IP address to initiate outgoing traffic)
1133 I.e. make sure that you can "ping" both the domain controller and also
1134 all of the clients from the node BEFORE you enable LVS. Also make sure
1135 that when you ping these hosts that the traffic is routed out through the
1141 <refsect1><title>REMOTE CLUSTER NODES</title>
1143 It is possible to have a CTDB cluster that spans across a WAN link.
1144 For example where you have a CTDB cluster in your datacentre but you also
1145 want to have one additional CTDB node located at a remote branch site.
1146 This is similar to how a WAN accelerator works but with the difference
1147 that while a WAN-accelerator often acts as a Proxy or a MitM, in
1148 the ctdb remote cluster node configuration the Samba instance at the remote site
1149 IS the genuine server, not a proxy and not a MitM, and thus provides 100%
1150 correct CIFS semantics to clients.
1154 See the cluster as one single multihomed samba server where one of
1155 the NICs (the remote node) is very far away.
1159 NOTE: This does require that the cluster filesystem you use can cope
1160 with WAN-link latencies. Not all cluster filesystems can handle
1161 WAN-link latencies! Whether this will provide very good WAN-accelerator
1162 performance or it will perform very poorly depends entirely
1163 on how optimized your cluster filesystem is in handling high latency
1164 for data and metadata operations.
1168 To activate a node as being a remote cluster node you need to set
1169 the following two parameters in /etc/sysconfig/ctdb for the remote node:
1170 <screen format="linespecific">
1171 CTDB_CAPABILITY_LMASTER=no
1172 CTDB_CAPABILITY_RECMASTER=no
1177 Verify with the command "ctdb getcapabilities" that that node no longer
1178 has the recmaster or the lmaster capabilities.
1184 <refsect1><title>NAT-GW</title>
1186 Sometimes it is desireable to run services on the CTDB node which will
1187 need to originate outgoing traffic to external servers. This might
1188 be contacting NIS servers, LDAP servers etc. etc.
1191 This can sometimes be problematic since there are situations when a
1192 node does not have any public ip addresses assigned. This could
1193 be due to the nobe just being started up and no addresses have been
1194 assigned yet or it could be that the node is UNHEALTHY in which
1195 case all public addresses have been migrated off.
1198 If then the service status of CTDB depends on such services being
1199 able to always being able to originate traffic to external resources
1200 this becomes extra troublesome. The node might be UNHEALTHY because
1201 the service can not be reached, and the service can not be reached
1202 because the node is UNHEALTHY.
1205 There are two ways to solve this problem. The first is by assigning a
1206 static ip address for one public interface on every node which will allow
1207 every node to be able to route traffic to the public network even
1208 if there are no public addresses assigned to the node.
1209 This is the simplest way but it uses up a lot of ip addresses since you
1210 have to assign both static and also public addresses to each node.
1212 <refsect2><title>NAT-GW</title>
1214 A second way is to use the built in NAT-GW feature in CTDB.
1215 With NAT-GW you assign one public NATGW address for each natgw group.
1216 Each NATGW group is a set of nodes in the cluster that shares the same
1217 NATGW address to talk to the outside world. Normally there would only be
1218 one NATGW group spanning the entire cluster, but in situations where one
1219 ctdb cluster spans multiple physical sites it is useful to have one
1220 NATGW group for each of the two sites.
1223 There can be multiple NATGW groups in one cluster but each node can only
1224 be member of one NATGW group.
1227 In each NATGW group, one of the nodes is designated the NAT Gateway
1228 through which all traffic that is originated by nodes in this group
1229 will be routed through if a public addresses are not available.
1233 <refsect2><title>Configuration</title>
1235 NAT-GW is configured in /etc/sysconfigctdb by setting the following
1238 <screen format="linespecific">
1239 # NAT-GW configuration
1240 # Some services running on nthe CTDB node may need to originate traffic to
1241 # remote servers before the node is assigned any IP addresses,
1242 # This is problematic since before the node has public addresses the node might
1243 # not be able to route traffic to the public networks.
1244 # One solution is to have static public addresses assigned with routing
1245 # in addition to the public address interfaces, thus guaranteeing that
1246 # a node always can route traffic to the external network.
1247 # This is the most simple solution but it uses up a large number of
1248 # additional ip addresses.
1250 # A more complex solution is NAT-GW.
1251 # In this mode we only need one additional ip address for the cluster from
1252 # the exsternal public network.
1253 # One of the nodes in the cluster is elected to be hosting this ip address
1254 # so it can reach the external services. This node is also configured
1255 # to use NAT MASQUERADING for all traffic from the internal private network
1256 # to the external network. This node is the NAT-GW node.
1258 # All other nodes are set up with a default rote with a metric of 10 to point
1259 # to the nat-gw node.
1261 # The effect of this is that only when a node does not have a public address
1262 # and thus no proper routes to the external world it will instead
1263 # route all packets through the nat-gw node.
1265 # CTDB_NATGW_NODES is the list of nodes that belong to this natgw group.
1266 # You can have multiple natgw groups in one cluster but each node
1267 # can only belong to one single natgw group.
1269 # CTDB_NATGW_PUBLIC_IP=10.0.0.227/24
1270 # CTDB_NATGW_PUBLIC_IFACE=eth0
1271 # CTDB_NATGW_DEFAULT_GATEWAY=10.0.0.1
1272 # CTDB_NATGW_PRIVATE_NETWORK=10.1.1.0/24
1273 # CTDB_NATGW_NODES=/etc/ctdb/natgw_nodes
1275 # Normally any node in the natgw group can act as the natgw master.
1276 # In some configurations you may have special nodes that is a part of the
1277 # cluster/natgw group, but where the node lacks connectivity to the
1279 # For these cases, set this variable to make these nodes not able to
1280 # become natgw master.
1282 # CTDB_NATGW_SLAVE_ONLY=yes
1286 <refsect2><title>CTDB_NATGW_PUBLIC_IP</title>
1288 This is an ip address in the public network that is used for all outgoing
1289 traffic when the public addresses are not assigned.
1290 This address will be assigned to one of the nodes in the cluster which
1291 will masquerade all traffic for the other nodes.
1294 Format of this parameter is IPADDRESS/NETMASK
1298 <refsect2><title>CTDB_NATGW_PUBLIC_IFACE</title>
1300 This is the physical interface where the CTDB_NATGW_PUBLIC_IP will be
1301 assigned to. This should be an interface connected to the public network.
1304 Format of this parameter is INTERFACE
1308 <refsect2><title>CTDB_NATGW_DEFAULT_GATEWAY</title>
1310 This is the default gateway to use on the node that is elected to host
1311 the CTDB_NATGW_PUBLIC_IP. This is the default gateway on the public network.
1314 Format of this parameter is IPADDRESS
1318 <refsect2><title>CTDB_NATGW_PRIVATE_NETWORK</title>
1320 This is the network/netmask used for the interal private network.
1323 Format of this parameter is IPADDRESS/NETMASK
1327 <refsect2><title>CTDB_NATGW_NODES</title>
1329 This is the list of all nodes that belong to the same NATGW group
1330 as this node. The default is /etc/ctdb/natgw_nodes.
1334 <refsect2><title>Operation</title>
1336 When the NAT-GW functionality is used, one of the nodes is elected
1337 to act as a NAT router for all the other nodes in the group when
1338 they need to originate traffic to the external public network.
1341 The NAT-GW node is assigned the CTDB_NATGW_PUBLIC_IP to the designated
1342 interface and the provided default route. The NAT-GW is configured
1343 to act as a router and to masquerade all traffic it receives from the
1344 internal private network and which is destined to the external network(s).
1347 All other nodes in the group are configured with a default route of
1348 metric 10 pointing to the designated NAT GW node.
1351 This is implemented in the 11.natgw eventscript. Please see the
1352 eventscript for further information.
1357 <refsect2><title>Removing/Changing NATGW at runtime</title>
1359 The following are the procedures to change/remove a NATGW configuration
1360 at runtime, without having to restart ctdbd.
1364 If you want to remove NATGW completely from a node, use these steps:
1366 <screen format="linespecific">
1367 1, Run 'CTDB_BASE=/etc/ctdb /etc/ctdb/events.d/11.natgw removenatgw'
1368 2, Then remove the configuration from /etc/sysconfig/ctdb
1372 If you want to change the NATGW configuration on a node :
1374 <screen format="linespecific">
1375 1, Run 'CTDB_BASE=/etc/ctdb /etc/ctdb/events.d/11.natgw removenatgw'
1376 2, Then change the configuration in /etc/sysconfig/ctdb
1377 3, Run 'CTDB_BASE=/etc/ctdb /etc/ctdb/events.d/11.natgw updatenatgw'
1384 <refsect1><title>NOTIFICATION SCRIPT</title>
1386 Notification scripts are used with ctdb to have a call-out from ctdb
1387 to a user-specified script when certain state changes occur in ctdb.
1388 This is commonly to set up either sending SNMP traps or emails
1389 when a node becomes unhealthy and similar.
1392 This is activated by setting CTDB_NOTIFY_SCRIPT=<your script> in the
1393 sysconfig file, or by adding --notification-script=<your script>.
1396 See /etc/ctdb/notify.sh for an example script.
1399 CTDB currently generates notifications on these state changes:
1402 <refsect2><title>unhealthy</title>
1404 This call-out is triggered when the node changes to UNHEALTHY state.
1408 <refsect2><title>healthy</title>
1410 This call-out is triggered when the node changes to HEALTHY state.
1414 <refsect2><title>startup</title>
1416 This call-out is triggered when ctdb has started up and all managed services are up and running.
1423 <refsect1><title>ClamAV Daemon</title>
1425 CTDB has support to manage the popular anti-virus daemon ClamAV.
1426 This support is implemented through the
1427 eventscript : /etc/ctdb/events.d/31.clamd.
1430 <refsect2><title>Configuration</title>
1432 Start by configuring CLAMAV normally and test that it works. Once this is
1433 done, copy the configuration files over to all the nodes so that all nodes
1434 share identical CLAMAV configurations.
1435 Once this is done you can proceed with the intructions below to activate
1436 CTDB support for CLAMAV.
1440 First, to activate CLAMAV support in CTDB, edit /etc/sysconfig/ctdb and add the two lines :
1442 <screen format="linespecific">
1443 CTDB_MANAGES_CLAMD=yes
1444 CTDB_CLAMD_SOCKET="/path/to/clamd.socket"
1448 Second, activate the eventscript
1450 <screen format="linespecific">
1451 ctdb enablescript 31.clamd
1455 Third, CTDB will now be starting and stopping this service accordingly,
1456 so make sure that the system is not configured to start/stop this service
1458 On RedHat systems you can disable the system starting/stopping CLAMAV automatically by running :
1459 <screen format="linespecific">
1466 Once you have restarted CTDBD, use
1467 <screen format="linespecific">
1470 and verify that the 31.clamd eventscript is listed and that it was executed successfully.
1479 <refsect1><title>SEE ALSO</title>
1482 <ulink url="http://ctdb.samba.org/"/>
1486 <refsect1><title>COPYRIGHT/LICENSE</title>
1488 Copyright (C) Andrew Tridgell 2007
1489 Copyright (C) Ronnie sahlberg 2007
1491 This program is free software; you can redistribute it and/or modify
1492 it under the terms of the GNU General Public License as published by
1493 the Free Software Foundation; either version 3 of the License, or (at
1494 your option) any later version.
1496 This program is distributed in the hope that it will be useful, but
1497 WITHOUT ANY WARRANTY; without even the implied warranty of
1498 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
1499 General Public License for more details.
1501 You should have received a copy of the GNU General Public License
1502 along with this program; if not, see http://www.gnu.org/licenses/.