1CTDB-TUNABLES(7) CTDB - clustered TDB database CTDB-TUNABLES(7)
2
6 ctdb-tunables - CTDB tunable configuration variables
7
9 CTDB's behaviour can be configured by setting run-time tunable
10 variables. This lists and describes all tunables. See the ctdb(1)
11 listvars, setvar and getvar commands for more details.
12 Unless otherwise stated, tunables should be set to the same value on
14 all nodes. Setting tunables to different values across nodes may
15 produce unexpected results. Future releases may set (some or most)
16 tunables globally across the cluster but doing so is currently a manual
17 process.
18 Tunables can be set at startup from the /etc/ctdb/ctdb.tunables
20 configuration file.
21 TUNABLE=VALUE
23 Comment lines beginning with '#' are permitted. Whitespace may be used
26 for formatting/alignment. VALUE must be a non-negative integer and must
27 be the last thing on a line (i.e. no trailing garbage, trailing
28 comments are not permitted).
29 For example:
31 MonitorInterval=20
33 The available tunable variables are listed alphabetically below.
36 AllowClientDBAttach
38 Default: 1
39 When set to 0, clients are not allowed to attach to any databases. This
41 can be used to temporarily block any new processes from attaching to
42 and accessing the databases. This is mainly used for detaching a
43 volatile database using 'ctdb detach'.
44 AllowMixedVersions
46 Default: 0
47 CTDB will not allow incompatible versions to co-exist in a cluster. If
49 a version mismatch is found, then losing CTDB will shutdown. To disable
50 the incompatible version check, set this tunable to 1.
51 For version checking, CTDB uses major and minor version. For example,
53 CTDB 4.6.1 and CTDB 4.6.2 are matching versions; CTDB 4.5.x and CTDB
54 4.6.y do not match.
55 CTDB with version check support will lose to CTDB without version check
57 support. Between two different CTDB versions with version check
58 support, one running for less time will lose. If the running time for
59 both CTDB versions with version check support is equal (to seconds),
60 then the older version will lose. The losing CTDB daemon will shutdown.
61 AllowUnhealthyDBRead
63 Default: 0
64 When set to 1, ctdb allows database traverses to read unhealthy
66 databases. By default, ctdb does not allow reading records from
67 unhealthy databases.
68 ControlTimeout
70 Default: 60
71 This is the default setting for timeout for when sending a control
73 message to either the local or a remote ctdb daemon.
74 DatabaseHashSize
76 Default: 100001
77 Number of the hash chains for the local store of the tdbs that ctdb
79 manages.
80 DatabaseMaxDead
82 Default: 5
83 Maximum number of dead records per hash chain for the tdb databases
85 managed by ctdb.
86 DBRecordCountWarn
88 Default: 100000
89 When set to non-zero, ctdb will log a warning during recovery if a
91 database has more than this many records. This will produce a warning
92 if a database grows uncontrollably with orphaned records.
93 DBRecordSizeWarn
95 Default: 10000000
96 When set to non-zero, ctdb will log a warning during recovery if a
98 single record is bigger than this size. This will produce a warning if
99 a database record grows uncontrollably.
100 DBSizeWarn
102 Default: 1000000000
103 When set to non-zero, ctdb will log a warning during recovery if a
105 database size is bigger than this. This will produce a warning if a
106 database grows uncontrollably.
107 DeferredAttachTO
109 Default: 120
110 When databases are frozen we do not allow clients to attach to the
112 databases. Instead of returning an error immediately to the client, the
113 attach request from the client is deferred until the database becomes
114 available again at which stage we respond to the client.
115 This timeout controls how long we will defer the request from the
117 client before timing it out and returning an error to the client.
118 ElectionTimeout
120 Default: 3
121 The number of seconds to wait for the election of recovery master to
123 complete. If the election is not completed during this interval, then
124 that round of election fails and ctdb starts a new election.
125 EnableBans
127 Default: 1
128 This parameter allows ctdb to ban a node if the node is misbehaving.
130 When set to 0, this disables banning completely in the cluster and thus
132 nodes can not get banned, even it they break. Don't set to 0 unless you
133 know what you are doing.
134 EventScriptTimeout
136 Default: 30
137 Maximum time in seconds to allow an event to run before timing out.
139 This is the total time for all enabled scripts that are run for an
140 event, not just a single event script.
141 Note that timeouts are ignored for some events ("takeip", "releaseip",
143 "startrecovery", "recovered") and converted to success. The logic here
144 is that the callers of these events implement their own additional
145 timeout.
146 FetchCollapse
148 Default: 1
149 This parameter is used to avoid multiple migration requests for the
151 same record from a single node. All the record requests for the same
152 record are queued up and processed when the record is migrated to the
153 current node.
154 When many clients across many nodes try to access the same record at
156 the same time this can lead to a fetch storm where the record becomes
157 very active and bounces between nodes very fast. This leads to high CPU
158 utilization of the ctdbd daemon, trying to bounce that record around
159 very fast, and poor performance. This can improve performance and
160 reduce CPU utilization for certain workloads.
161 HopcountMakeSticky
163 Default: 50
164 For database(s) marked STICKY (using 'ctdb setdbsticky'), any record
166 that is migrating so fast that hopcount exceeds this limit is marked as
167 STICKY record for StickyDuration seconds. This means that after each
168 migration the sticky record will be kept on the node
169 StickyPindownmilliseconds and prevented from being migrated off the
170 node.
171 This will improve performance for certain workloads, such as
173 locking.tdb if many clients are opening/closing the same file
174 concurrently.
175 IPAllocAlgorithm
177 Default: 2
178 Selects the algorithm that CTDB should use when doing public IP address
180 allocation. Meaningful values are:
181 0
183 Deterministic IP address allocation.
184 This is a simple and fast option. However, it can cause unnecessary
186 address movement during fail-over because each address has a "home"
187 node. Works badly when some nodes do not have any addresses
188 defined. Should be used with care when addresses are defined across
189 multiple networks.
190 1
192 Non-deterministic IP address allocation.
193 This is a relatively fast option that attempts to do a minimise
195 unnecessary address movements. Addresses do not have a "home" node.
196 Rebalancing is limited but it usually adequate. Works badly when
197 addresses are defined across multiple networks.
198 2
200 LCP2 IP address allocation.
201 Uses a heuristic to assign addresses defined across multiple
203 networks, usually balancing addresses on each network evenly across
204 nodes. Addresses do not have a "home" node. Minimises unnecessary
205 address movements. The algorithm is complex, so is slower than
206 other choices for a large number of addresses. However, it can
207 calculate an optimal assignment of 900 addresses in under 10
208 seconds on modern hardware.
209 If the specified value is not one of these then the default will be
211 used.
212 KeepaliveInterval
214 Default: 5
215 How often in seconds should the nodes send keep-alive packets to each
217 other.
218 KeepaliveLimit
220 Default: 5
221 After how many keepalive intervals without any traffic should a node
223 wait until marking the peer as DISCONNECTED.
224 If a node has hung, it can take KeepaliveInterval * (KeepaliveLimit +
226 1) seconds before ctdb determines that the node is DISCONNECTED and
227 performs a recovery. This limit should not be set too high to enable
228 early detection and avoid any application timeouts (e.g. SMB1) to kick
229 in before the fail over is completed.
230 LockProcessesPerDB
232 Default: 200
233 This is the maximum number of lock helper processes ctdb will create
235 for obtaining record locks. When ctdb cannot get a record lock without
236 blocking, it creates a helper process that waits for the lock to be
237 obtained.
238 LogLatencyMs
240 Default: 0
241 When set to non-zero, ctdb will log if certains operations take longer
243 than this value, in milliseconds, to complete. These operations include
244 "process a record request from client", "take a record or database
245 lock", "update a persistent database record" and "vacuum a database".
246 MaxQueueDropMsg
248 Default: 1000000
249 This is the maximum number of messages to be queued up for a client
251 before ctdb will treat the client as hung and will terminate the client
252 connection.
253 MonitorInterval
255 Default: 15
256 How often should ctdb run the 'monitor' event in seconds to check for a
258 node's health.
259 MonitorTimeoutCount
261 Default: 20
262 How many 'monitor' events in a row need to timeout before a node is
264 flagged as UNHEALTHY. This setting is useful if scripts can not be
265 written so that they do not hang for benign reasons.
266 NoIPFailback
268 Default: 0
269 When set to 1, ctdb will not perform failback of IP addresses when a
271 node becomes healthy. When a node becomes UNHEALTHY, ctdb WILL perform
272 failover of public IP addresses, but when the node becomes HEALTHY
273 again, ctdb will not fail the addresses back.
274 Use with caution! Normally when a node becomes available to the cluster
276 ctdb will try to reassign public IP addresses onto the new node as a
277 way to distribute the workload evenly across the clusternode. Ctdb
278 tries to make sure that all running nodes have approximately the same
279 number of public addresses it hosts.
280 When you enable this tunable, ctdb will no longer attempt to rebalance
282 the cluster by failing IP addresses back to the new nodes. An
283 unbalanced cluster will therefore remain unbalanced until there is
284 manual intervention from the administrator. When this parameter is set,
285 you can manually fail public IP addresses over to the new node(s) using
286 the 'ctdb moveip' command.
287 NoIPTakeover
289 Default: 0
290 When set to 1, ctdb will not allow IP addresses to be failed over to
292 other nodes. Any IP addresses already hosted on healthy nodes will
293 remain. Any IP addresses hosted on unhealthy nodes will be released by
294 unhealthy nodes and will become un-hosted.
295 PullDBPreallocation
297 Default: 10*1024*1024
298 This is the size of a record buffer to pre-allocate for sending reply
300 to PULLDB control. Usually record buffer starts with size of the first
301 record and gets reallocated every time a new record is added to the
302 record buffer. For a large number of records, this can be very
303 inefficient to grow the record buffer one record at a time.
304 QueueBufferSize
306 Default: 1024
307 This is the maximum amount of data (in bytes) ctdb will read from a
309 socket at a time.
310 For a busy setup, if ctdb is not able to process the TCP sockets fast
312 enough (large amount of data in Recv-Q for tcp sockets), then this
313 tunable value should be increased. However, large values can keep ctdb
314 busy processing packets and prevent ctdb from handling other events.
315 RecBufferSizeLimit
317 Default: 1000000
318 This is the limit on the size of the record buffer to be sent in
320 various controls. This limit is used by new controls used for recovery
321 and controls used in vacuuming.
322 RecdFailCount
324 Default: 10
325 If the recovery daemon has failed to ping the main daemon for this many
327 consecutive intervals, the main daemon will consider the recovery
328 daemon as hung and will try to restart it to recover.
329 RecdPingTimeout
331 Default: 60
332 If the main daemon has not heard a "ping" from the recovery daemon for
334 this many seconds, the main daemon will log a message that the recovery
335 daemon is potentially hung. This also increments a counter which is
336 checked against RecdFailCount for detection of hung recovery daemon.
337 RecLockLatencyMs
339 Default: 1000
340 When using a reclock file for split brain prevention, if set to
342 non-zero this tunable will make the recovery daemon log a message if
343 the fcntl() call to lock/testlock the recovery file takes longer than
344 this number of milliseconds.
345 RecoverInterval
347 Default: 1
348 How frequently in seconds should the recovery daemon perform the
350 consistency checks to determine if it should perform a recovery.
351 RecoverTimeout
353 Default: 120
354 This is the default setting for timeouts for controls when sent from
356 the recovery daemon. We allow longer control timeouts from the recovery
357 daemon than from normal use since the recovery daemon often use
358 controls that can take a lot longer than normal controls.
359 RecoveryBanPeriod
361 Default: 300
362 The duration in seconds for which a node is banned if the node fails
364 during recovery. After this time has elapsed the node will
365 automatically get unbanned and will attempt to rejoin the cluster.
366 A node usually gets banned due to real problems with the node. Don't
368 set this value too small. Otherwise, a problematic node will try to
369 re-join cluster too soon causing unnecessary recoveries.
370 RecoveryDropAllIPs
372 Default: 120
373 If a node is stuck in recovery, or stopped, or banned, for this many
375 seconds, then ctdb will release all public addresses on that node.
376 RecoveryGracePeriod
378 Default: 120
379 During recoveries, if a node has not caused recovery failures during
381 the last grace period in seconds, any records of transgressions that
382 the node has caused recovery failures will be forgiven. This resets the
383 ban-counter back to zero for that node.
384 RepackLimit
386 Default: 10000
387 During vacuuming, if the number of freelist records are more than
389 RepackLimit, then the database is repacked to get rid of the freelist
390 records to avoid fragmentation.
391 RerecoveryTimeout
393 Default: 10
394 Once a recovery has completed, no additional recoveries are permitted
396 until this timeout in seconds has expired.
397 SeqnumInterval
399 Default: 1000
400 Some databases have seqnum tracking enabled, so that samba will be able
402 to detect asynchronously when there has been updates to the database.
403 Every time a database is updated its sequence number is increased.
404 This tunable is used to specify in milliseconds how frequently ctdb
406 will send out updates to remote nodes to inform them that the sequence
407 number is increased.
408 StatHistoryInterval
410 Default: 1
411 Granularity of the statistics collected in the statistics history. This
413 is reported by 'ctdb stats' command.
414 StickyDuration
416 Default: 600
417 Once a record has been marked STICKY, this is the duration in seconds,
419 the record will be flagged as a STICKY record.
420 StickyPindown
422 Default: 200
423 Once a STICKY record has been migrated onto a node, it will be pinned
425 down on that node for this number of milliseconds. Any request from
426 other nodes to migrate the record off the node will be deferred.
427 TakeoverTimeout
429 Default: 9
430 This is the duration in seconds in which ctdb tries to complete IP
432 failover.
433 TickleUpdateInterval
435 Default: 20
436 Every TickleUpdateInterval seconds, ctdb synchronizes the client
438 connection information across nodes.
439 TraverseTimeout
441 Default: 20
442 This is the duration in seconds for which a database traverse is
444 allowed to run. If the traverse does not complete during this interval,
445 ctdb will abort the traverse.
446 VacuumFastPathCount
448 Default: 60
449 During a vacuuming run, ctdb usually processes only the records marked
451 for deletion also called the fast path vacuuming. After finishing
452 VacuumFastPathCount number of fast path vacuuming runs, ctdb will
453 trigger a scan of complete database for any empty records that need to
454 be deleted.
455 VacuumInterval
457 Default: 10
458 Periodic interval in seconds when vacuuming is triggered for volatile
460 databases.
461 VacuumMaxRunTime
463 Default: 120
464 The maximum time in seconds for which the vacuuming process is allowed
466 to run. If vacuuming process takes longer than this value, then the
467 vacuuming process is terminated.
468 VerboseMemoryNames
470 Default: 0
471 When set to non-zero, ctdb assigns verbose names for some of the talloc
473 allocated memory objects. These names are visible in the talloc memory
474 report generated by 'ctdb dumpmemory'.
475
477 /etc/ctdb/ctdb.tunables
478
480 ctdb(1), ctdbd(1), ctdb.conf(5), ctdb(7), http://ctdb.samba.org/
481
483 This documentation was written by Ronnie Sahlberg, Amitay Isaacs,
484 Martin Schwenke
485
487 Copyright © 2007 Andrew Tridgell, Ronnie Sahlberg
488 This program is free software; you can redistribute it and/or modify it
490 under the terms of the GNU General Public License as published by the
491 Free Software Foundation; either version 3 of the License, or (at your
492 option) any later version.
493 This program is distributed in the hope that it will be useful, but
495 WITHOUT ANY WARRANTY; without even the implied warranty of
496 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
497 General Public License for more details.
498 You should have received a copy of the GNU General Public License along
500 with this program; if not, see http://www.gnu.org/licenses.
501ctdb 11/27/2023 CTDB-TUNABLES(7)