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 For example:
26 MonitorInterval=20
28 The available tunable variables are listed alphabetically below.
31 AllowClientDBAttach
33 Default: 1
34 When set to 0, clients are not allowed to attach to any databases. This
36 can be used to temporarily block any new processes from attaching to
37 and accessing the databases. This is mainly used for detaching a
38 volatile database using 'ctdb detach'.
39 AllowMixedVersions
41 Default: 0
42 CTDB will not allow incompatible versions to co-exist in a cluster. If
44 a version mismatch is found, then losing CTDB will shutdown. To disable
45 the incompatible version check, set this tunable to 1.
46 For version checking, CTDB uses major and minor version. For example,
48 CTDB 4.6.1 and CTDB CTDB 4.6.2 are matching versions; CTDB 4.5.x and
49 CTDB 4.6.y do not match.
50 CTDB with version check support will lose to CTDB without version check
52 support. Between two different CTDB versions with version check
53 support, one running for less time will lose. If the running time for
54 both CTDB versions with version check support is equal (to seconds),
55 then the older version will lose. The losing CTDB daemon will shutdown.
56 AllowUnhealthyDBRead
58 Default: 0
59 When set to 1, ctdb allows database traverses to read unhealthy
61 databases. By default, ctdb does not allow reading records from
62 unhealthy databases.
63 ControlTimeout
65 Default: 60
66 This is the default setting for timeout for when sending a control
68 message to either the local or a remote ctdb daemon.
69 DatabaseHashSize
71 Default: 100001
72 Number of the hash chains for the local store of the tdbs that ctdb
74 manages.
75 DatabaseMaxDead
77 Default: 5
78 Maximum number of dead records per hash chain for the tdb databses
80 managed by ctdb.
81 DBRecordCountWarn
83 Default: 100000
84 When set to non-zero, ctdb will log a warning during recovery if a
86 database has more than this many records. This will produce a warning
87 if a database grows uncontrollably with orphaned records.
88 DBRecordSizeWarn
90 Default: 10000000
91 When set to non-zero, ctdb will log a warning during recovery if a
93 single record is bigger than this size. This will produce a warning if
94 a database record grows uncontrollably.
95 DBSizeWarn
97 Default: 1000000000
98 When set to non-zero, ctdb will log a warning during recovery if a
100 database size is bigger than this. This will produce a warning if a
101 database grows uncontrollably.
102 DeferredAttachTO
104 Default: 120
105 When databases are frozen we do not allow clients to attach to the
107 databases. Instead of returning an error immediately to the client, the
108 attach request from the client is deferred until the database becomes
109 available again at which stage we respond to the client.
110 This timeout controls how long we will defer the request from the
112 client before timing it out and returning an error to the client.
113 ElectionTimeout
115 Default: 3
116 The number of seconds to wait for the election of recovery master to
118 complete. If the election is not completed during this interval, then
119 that round of election fails and ctdb starts a new election.
120 EnableBans
122 Default: 1
123 This parameter allows ctdb to ban a node if the node is misbehaving.
125 When set to 0, this disables banning completely in the cluster and thus
127 nodes can not get banned, even it they break. Don't set to 0 unless you
128 know what you are doing.
129 EventScriptTimeout
131 Default: 30
132 Maximum time in seconds to allow an event to run before timing out.
134 This is the total time for all enabled scripts that are run for an
135 event, not just a single event script.
136 Note that timeouts are ignored for some events ("takeip", "releaseip",
138 "startrecovery", "recovered") and converted to success. The logic here
139 is that the callers of these events implement their own additional
140 timeout.
141 FetchCollapse
143 Default: 1
144 This parameter is used to avoid multiple migration requests for the
146 same record from a single node. All the record requests for the same
147 record are queued up and processed when the record is migrated to the
148 current node.
149 When many clients across many nodes try to access the same record at
151 the same time this can lead to a fetch storm where the record becomes
152 very active and bounces between nodes very fast. This leads to high CPU
153 utilization of the ctdbd daemon, trying to bounce that record around
154 very fast, and poor performance. This can improve performance and
155 reduce CPU utilization for certain workloads.
156 HopcountMakeSticky
158 Default: 50
159 For database(s) marked STICKY (using 'ctdb setdbsticky'), any record
161 that is migrating so fast that hopcount exceeds this limit is marked as
162 STICKY record for StickyDuration seconds. This means that after each
163 migration the sticky record will be kept on the node
164 StickyPindownmilliseconds and prevented from being migrated off the
165 node.
166 This will improve performance for certain workloads, such as
168 locking.tdb if many clients are opening/closing the same file
169 concurrently.
170 IPAllocAlgorithm
172 Default: 2
173 Selects the algorithm that CTDB should use when doing public IP address
175 allocation. Meaningful values are:
176 0
178 Deterministic IP address allocation.
179 This is a simple and fast option. However, it can cause unnecessary
181 address movement during fail-over because each address has a "home"
182 node. Works badly when some nodes do not have any addresses
183 defined. Should be used with care when addresses are defined across
184 multiple networks.
185 1
187 Non-deterministic IP address allocation.
188 This is a relatively fast option that attempts to do a minimise
190 unnecessary address movements. Addresses do not have a "home" node.
191 Rebalancing is limited but it usually adequate. Works badly when
192 addresses are defined across multiple networks.
193 2
195 LCP2 IP address allocation.
196 Uses a heuristic to assign addresses defined across multiple
198 networks, usually balancing addresses on each network evenly across
199 nodes. Addresses do not have a "home" node. Minimises unnecessary
200 address movements. The algorithm is complex, so is slower than
201 other choices for a large number of addresses. However, it can
202 calculate an optimal assignment of 900 addresses in under 10
203 seconds on modern hardware.
204 If the specified value is not one of these then the default will be
206 used.
207 KeepaliveInterval
209 Default: 5
210 How often in seconds should the nodes send keep-alive packets to each
212 other.
213 KeepaliveLimit
215 Default: 5
216 After how many keepalive intervals without any traffic should a node
218 wait until marking the peer as DISCONNECTED.
219 If a node has hung, it can take KeepaliveInterval * (KeepaliveLimit +
221 1) seconds before ctdb determines that the node is DISCONNECTED and
222 performs a recovery. This limit should not be set too high to enable
223 early detection and avoid any application timeouts (e.g. SMB1) to kick
224 in before the fail over is completed.
225 LockProcessesPerDB
227 Default: 200
228 This is the maximum number of lock helper processes ctdb will create
230 for obtaining record locks. When ctdb cannot get a record lock without
231 blocking, it creates a helper process that waits for the lock to be
232 obtained.
233 LogLatencyMs
235 Default: 0
236 When set to non-zero, ctdb will log if certains operations take longer
238 than this value, in milliseconds, to complete. These operations include
239 "process a record request from client", "take a record or database
240 lock", "update a persistent database record" and "vaccum a database".
241 MaxQueueDropMsg
243 Default: 1000000
244 This is the maximum number of messages to be queued up for a client
246 before ctdb will treat the client as hung and will terminate the client
247 connection.
248 MonitorInterval
250 Default: 15
251 How often should ctdb run the 'monitor' event in seconds to check for a
253 node's health.
254 MonitorTimeoutCount
256 Default: 20
257 How many 'monitor' events in a row need to timeout before a node is
259 flagged as UNHEALTHY. This setting is useful if scripts can not be
260 written so that they do not hang for benign reasons.
261 NoIPFailback
263 Default: 0
264 When set to 1, ctdb will not perform failback of IP addresses when a
266 node becomes healthy. When a node becomes UNHEALTHY, ctdb WILL perform
267 failover of public IP addresses, but when the node becomes HEALTHY
268 again, ctdb will not fail the addresses back.
269 Use with caution! Normally when a node becomes available to the cluster
271 ctdb will try to reassign public IP addresses onto the new node as a
272 way to distribute the workload evenly across the clusternode. Ctdb
273 tries to make sure that all running nodes have approximately the same
274 number of public addresses it hosts.
275 When you enable this tunable, ctdb will no longer attempt to rebalance
277 the cluster by failing IP addresses back to the new nodes. An
278 unbalanced cluster will therefore remain unbalanced until there is
279 manual intervention from the administrator. When this parameter is set,
280 you can manually fail public IP addresses over to the new node(s) using
281 the 'ctdb moveip' command.
282 NoIPTakeover
284 Default: 0
285 When set to 1, ctdb will not allow IP addresses to be failed over to
287 other nodes. Any IP addresses already hosted on healthy nodes will
288 remain. Any IP addresses hosted on unhealthy nodes will be released by
289 unhealthy nodes and will become un-hosted.
290 PullDBPreallocation
292 Default: 10*1024*1024
293 This is the size of a record buffer to pre-allocate for sending reply
295 to PULLDB control. Usually record buffer starts with size of the first
296 record and gets reallocated every time a new record is added to the
297 record buffer. For a large number of records, this can be very
298 inefficient to grow the record buffer one record at a time.
299 QueueBufferSize
301 Default: 1024
302 This is the maximum amount of data (in bytes) ctdb will read from a
304 socket at a time.
305 For a busy setup, if ctdb is not able to process the TCP sockets fast
307 enough (large amount of data in Recv-Q for tcp sockets), then this
308 tunable value should be increased. However, large values can keep ctdb
309 busy processing packets and prevent ctdb from handling other events.
310 RecBufferSizeLimit
312 Default: 1000000
313 This is the limit on the size of the record buffer to be sent in
315 various controls. This limit is used by new controls used for recovery
316 and controls used in vacuuming.
317 RecdFailCount
319 Default: 10
320 If the recovery daemon has failed to ping the main dameon for this many
322 consecutive intervals, the main daemon will consider the recovery
323 daemon as hung and will try to restart it to recover.
324 RecdPingTimeout
326 Default: 60
327 If the main dameon has not heard a "ping" from the recovery dameon for
329 this many seconds, the main dameon will log a message that the recovery
330 daemon is potentially hung. This also increments a counter which is
331 checked against RecdFailCount for detection of hung recovery daemon.
332 RecLockLatencyMs
334 Default: 1000
335 When using a reclock file for split brain prevention, if set to
337 non-zero this tunable will make the recovery dameon log a message if
338 the fcntl() call to lock/testlock the recovery file takes longer than
339 this number of milliseconds.
340 RecoverInterval
342 Default: 1
343 How frequently in seconds should the recovery daemon perform the
345 consistency checks to determine if it should perform a recovery.
346 RecoverTimeout
348 Default: 120
349 This is the default setting for timeouts for controls when sent from
351 the recovery daemon. We allow longer control timeouts from the recovery
352 daemon than from normal use since the recovery dameon often use
353 controls that can take a lot longer than normal controls.
354 RecoveryBanPeriod
356 Default: 300
357 The duration in seconds for which a node is banned if the node fails
359 during recovery. After this time has elapsed the node will
360 automatically get unbanned and will attempt to rejoin the cluster.
361 A node usually gets banned due to real problems with the node. Don't
363 set this value too small. Otherwise, a problematic node will try to
364 re-join cluster too soon causing unnecessary recoveries.
365 RecoveryDropAllIPs
367 Default: 120
368 If a node is stuck in recovery, or stopped, or banned, for this many
370 seconds, then ctdb will release all public addresses on that node.
371 RecoveryGracePeriod
373 Default: 120
374 During recoveries, if a node has not caused recovery failures during
376 the last grace period in seconds, any records of transgressions that
377 the node has caused recovery failures will be forgiven. This resets the
378 ban-counter back to zero for that node.
379 RepackLimit
381 Default: 10000
382 During vacuuming, if the number of freelist records are more than
384 RepackLimit, then the database is repacked to get rid of the freelist
385 records to avoid fragmentation.
386 Databases are repacked only if both RepackLimit and VacuumLimit are
388 exceeded.
389 RerecoveryTimeout
391 Default: 10
392 Once a recovery has completed, no additional recoveries are permitted
394 until this timeout in seconds has expired.
395 SeqnumInterval
397 Default: 1000
398 Some databases have seqnum tracking enabled, so that samba will be able
400 to detect asynchronously when there has been updates to the database.
401 Every time a database is updated its sequence number is increased.
402 This tunable is used to specify in milliseconds how frequently ctdb
404 will send out updates to remote nodes to inform them that the sequence
405 number is increased.
406 StatHistoryInterval
408 Default: 1
409 Granularity of the statistics collected in the statistics history. This
411 is reported by 'ctdb stats' command.
412 StickyDuration
414 Default: 600
415 Once a record has been marked STICKY, this is the duration in seconds,
417 the record will be flagged as a STICKY record.
418 StickyPindown
420 Default: 200
421 Once a STICKY record has been migrated onto a node, it will be pinned
423 down on that node for this number of milliseconds. Any request from
424 other nodes to migrate the record off the node will be deferred.
425 TakeoverTimeout
427 Default: 9
428 This is the duration in seconds in which ctdb tries to complete IP
430 failover.
431 TickleUpdateInterval
433 Default: 20
434 Every TickleUpdateInterval seconds, ctdb synchronizes the client
436 connection information across nodes.
437 TraverseTimeout
439 Default: 20
440 This is the duration in seconds for which a database traverse is
442 allowed to run. If the traverse does not complete during this interval,
443 ctdb will abort the traverse.
444 VacuumFastPathCount
446 Default: 60
447 During a vacuuming run, ctdb usually processes only the records marked
449 for deletion also called the fast path vacuuming. After finishing
450 VacuumFastPathCount number of fast path vacuuming runs, ctdb will
451 trigger a scan of complete database for any empty records that need to
452 be deleted.
453 VacuumInterval
455 Default: 10
456 Periodic interval in seconds when vacuuming is triggered for volatile
458 databases.
459 VacuumLimit
461 Default: 5000
462 During vacuuming, if the number of deleted records are more than
464 VacuumLimit, then databases are repacked to avoid fragmentation.
465 Databases are repacked only if both RepackLimit and VacuumLimit are
467 exceeded.
468 VacuumMaxRunTime
470 Default: 120
471 The maximum time in seconds for which the vacuuming process is allowed
473 to run. If vacuuming process takes longer than this value, then the
474 vacuuming process is terminated.
475 VerboseMemoryNames
477 Default: 0
478 When set to non-zero, ctdb assigns verbose names for some of the talloc
480 allocated memory objects. These names are visible in the talloc memory
481 report generated by 'ctdb dumpmemory'.
482
484 /etc/ctdb/ctdb.tunables
485
487 ctdb(1), ctdbd(1), ctdb.conf(5), ctdb(7), http://ctdb.samba.org/
488
490 This documentation was written by Ronnie Sahlberg, Amitay Isaacs,
491 Martin Schwenke
492
494 Copyright © 2007 Andrew Tridgell, Ronnie Sahlberg
495 This program is free software; you can redistribute it and/or modify it
497 under the terms of the GNU General Public License as published by the
498 Free Software Foundation; either version 3 of the License, or (at your
499 option) any later version.
500 This program is distributed in the hope that it will be useful, but
502 WITHOUT ANY WARRANTY; without even the implied warranty of
503 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
504 General Public License for more details.
505 You should have received a copy of the GNU General Public License along
507 with this program; if not, see http://www.gnu.org/licenses.
508ctdb 05/14/2019 CTDB-TUNABLES(7)