1DRBD.CONF(5) Configuration Files DRBD.CONF(5)
2
6 drbd.conf - Configuration file for DRBD's devices
7
9 The file /etc/drbd.conf is read by drbdadm.
10 The file format was designed as to allow to have a verbatim copy of the
12 file on both nodes of the cluster. It is highly recommended to do so in
13 order to keep your configuration manageable. The file /etc/drbd.conf
14 should be the same on both nodes of the cluster. Changes to
15 /etc/drbd.conf do not apply immediately.
16 By convention the main config contains two include statements. The
18 first one includes the file /etc/drbd.d/global_common.conf, the second
19 one all file with a .res suffix.
20 resource r0 {
22 net {
23 protocol C;
24 cram-hmac-alg sha1;
25 shared-secret "FooFunFactory";
26 }
27 disk {
28 resync-rate 10M;
29 }
30 on alice {
31 volume 0 {
32 device minor 1;
33 disk /dev/sda7;
34 meta-disk internal;
35 }
36 address 10.1.1.31:7789;
37 }
38 on bob {
39 volume 0 {
40 device minor 1;
41 disk /dev/sda7;
42 meta-disk internal;
43 }
44 address 10.1.1.32:7789;
45 }
46 }
47 In this example, there is a single DRBD resource (called r0) which uses
49 protocol C for the connection between its devices. It contains a single
50 volume which runs on host alice uses /dev/drbd1 as devices for its
51 application, and /dev/sda7 as low-level storage for the data. The IP
52 addresses are used to specify the networking interfaces to be used. An
53 eventually running resync process should use about 10MByte/second of IO
54 bandwidth. This resync-rate statement is valid for volume 0, but would
55 also be valid for further volumes. In this example it assigns full
56 10MByte/second to each volume.
57 There may be multiple resource sections in a single drbd.conf file. For
59 more examples, please have a look at the DRBD User's Guide[1].
60
62 The file consists of sections and parameters. A section begins with a
63 keyword, sometimes an additional name, and an opening brace (“{”). A
64 section ends with a closing brace (“}”. The braces enclose the
65 parameters.
66 section [name] { parameter value; [...] }
68 A parameter starts with the identifier of the parameter followed by
70 whitespace. Every subsequent character is considered as part of the
71 parameter's value. A special case are Boolean parameters which consist
72 only of the identifier. Parameters are terminated by a semicolon (“;”).
73 Some parameter values have default units which might be overruled by K,
75 M or G. These units are defined in the usual way (K = 2^10 = 1024, M =
76 1024 K, G = 1024 M).
77 Comments may be placed into the configuration file and must begin with
79 a hash sign (“#”). Subsequent characters are ignored until the end of
80 the line.
81 Sections
83 skip
84 Comments out chunks of text, even spanning more than one line.
86 Characters between the keyword skip and the opening brace (“{”) are
87 ignored. Everything enclosed by the braces is skipped. This comes
88 in handy, if you just want to comment out some 'resource [name]
89 {...}' section: just precede it with 'skip'.
90 global
92 Configures some global parameters. Currently only minor-count,
94 dialog-refresh, disable-ip-verification and usage-count are allowed
95 here. You may only have one global section, preferably as the first
96 section.
97 common
99 All resources inherit the options set in this section. The common
101 section might have a startup, a options, a handlers, a net and a
102 disk section.
103 resource name
105 Configures a DRBD resource. Each resource section needs to have two
107 (or more) on host sections and may have a startup, a options, a
108 handlers, a net and a disk section. It might contain volumes
109 sections.
110 on host-name
112 Carries the necessary configuration parameters for a DRBD device of
114 the enclosing resource. host-name is mandatory and must match the
115 Linux host name (uname -n) of one of the nodes. You may list more
116 than one host name here, in case you want to use the same
117 parameters on several hosts (you'd have to move the IP around
118 usually). Or you may list more than two such sections.
119 resource r1 {
121 protocol C;
122 device minor 1;
123 meta-disk internal;
124 on alice bob {
126 address 10.2.2.100:7801;
127 disk /dev/mapper/some-san;
128 }
129 on charlie {
130 address 10.2.2.101:7801;
131 disk /dev/mapper/other-san;
132 }
133 on daisy {
134 address 10.2.2.103:7801;
135 disk /dev/mapper/other-san-as-seen-from-daisy;
136 }
137 }
138 See also the floating section keyword. Required statements in this
141 section: address and volume. Note for backward compatibility and
142 convenience it is valid to embed the statements of a single volume
143 directly into the host section.
144 volume vnr
146 Defines a volume within a connection. The minor numbers of a
148 replicated volume might be different on different hosts, the volume
149 number (vnr) is what groups them together. Required parameters in
150 this section: device, disk, meta-disk.
151 stacked-on-top-of resource
153 For a stacked DRBD setup (3 or 4 nodes), a stacked-on-top-of is
155 used instead of an on section. Required parameters in this section:
156 device and address.
157 floating AF addr:port
159 Carries the necessary configuration parameters for a DRBD device of
161 the enclosing resource. This section is very similar to the on
162 section. The difference to the on section is that the matching of
163 the host sections to machines is done by the IP-address instead of
164 the node name. Required parameters in this section: device, disk,
165 meta-disk, all of which may be inherited from the resource section,
166 in which case you may shorten this section down to just the address
167 identifier.
168 resource r2 {
170 protocol C;
171 device minor 2;
172 disk /dev/sda7;
173 meta-disk internal;
174 # short form, device, disk and meta-disk inherited
176 floating 10.1.1.31:7802;
177 # longer form, only device inherited
179 floating 10.1.1.32:7802 {
180 disk /dev/sdb;
181 meta-disk /dev/sdc8;
182 }
183 }
184 disk
187 This section is used to fine tune DRBD's properties in respect to
189 the low level storage. Please refer to drbdsetup(8) for detailed
190 description of the parameters. Optional parameters: on-io-error,
191 size, fencing, disk-barrier, disk-flushes, disk-drain, md-flushes,
192 max-bio-bvecs, resync-rate, resync-after, al-extents, al-updates,
193 c-plan-ahead, c-fill-target, c-delay-target, c-max-rate,
194 c-min-rate, disk-timeout, discard-zeroes-if-aligned,
195 rs-discard-granularity, read-balancing.
196 net
198 This section is used to fine tune DRBD's properties. Please refer
200 to drbdsetup(8) for a detailed description of this section's
201 parameters. Optional parameters: protocol, sndbuf-size,
202 rcvbuf-size, timeout, connect-int, ping-int, ping-timeout,
203 max-buffers, max-epoch-size, ko-count, allow-two-primaries,
204 cram-hmac-alg, shared-secret, after-sb-0pri, after-sb-1pri,
205 after-sb-2pri, data-integrity-alg, no-tcp-cork, on-congestion,
206 congestion-fill, congestion-extents, verify-alg, use-rle,
207 csums-alg, socket-check-timeout.
208 startup
210 This section is used to fine tune DRBD's properties. Please refer
212 to drbdsetup(8) for a detailed description of this section's
213 parameters. Optional parameters: wfc-timeout, degr-wfc-timeout,
214 outdated-wfc-timeout, wait-after-sb, stacked-timeouts and
215 become-primary-on.
216 options
218 This section is used to fine tune the behaviour of the resource
220 object. Please refer to drbdsetup(8) for a detailed description of
221 this section's parameters. Optional parameters: cpu-mask, and
222 on-no-data-accessible.
223 handlers
225 In this section you can define handlers (executables) that are
227 started by the DRBD system in response to certain events. Optional
228 parameters: pri-on-incon-degr, pri-lost-after-sb, pri-lost,
229 fence-peer (formerly oudate-peer), local-io-error,
230 initial-split-brain, split-brain, before-resync-target,
231 after-resync-target.
232 The interface is done via environment variables:
234 • DRBD_RESOURCE is the name of the resource
236 • DRBD_MINOR is the minor number of the DRBD device, in decimal.
238 • DRBD_CONF is the path to the primary configuration file; if you
240 split your configuration into multiple files (e.g. in
241 /etc/drbd.conf.d/), this will not be helpful.
242 • DRBD_PEER_AF , DRBD_PEER_ADDRESS , DRBD_PEERS are the address
244 family (e.g. ipv6), the peer's address and hostnames.
245 DRBD_PEER is deprecated.
248 Please note that not all of these might be set for all handlers,
250 and that some values might not be useable for a floating
251 definition.
252 Parameters
254 minor-count count
255 count may be a number from 1 to 1048575.
256 Minor-count is a sizing hint for DRBD. It helps to right-size
258 various memory pools. It should be set in the in the same order of
259 magnitude than the actual number of minors you use. Per default the
260 module loads with 11 more resources than you have currently in your
261 config but at least 32.
262 dialog-refresh time
264 time may be 0 or a positive number.
265 The user dialog redraws the second count every time seconds (or
267 does no redraws if time is 0). The default value is 1.
268 disable-ip-verification
270 Use disable-ip-verification if, for some obscure reasons, drbdadm
271 can/might not use ip or ifconfig to do a sanity check for the IP
272 address. You can disable the IP verification with this option.
273 udev-always-use-vnr
275 When udev asks drbdadm for a list of device related symlinks,
276 drbdadm would suggest symlinks with differing naming conventions,
277 depending on whether the resource has explicit volume VNR { }
278 definitions, or only one single volume with the implicit volume
279 number 0:
280 # implicit single volume without "volume 0 {}" block
282 DEVICE=drbd<minor>
283 SYMLINK_BY_RES=drbd/by-res/<resource-name>
284 # explicit volume definition: volume VNR { }
286 DEVICE=drbd<minor>
287 SYMLINK_BY_RES=drbd/by-res/<resource-name>/VNR
288 If you define this parameter in the global section, drbdadm will
290 always add the .../VNR part, and will not care for whether the
291 volume definition was implicit or explicit.
292 For legacy backward compatibility, this is off by default, but we
294 do recommend to enable it.
295 usage-count val
297 Please participate in DRBD's online usage counter[2]. The most
298 convenient way to do so is to set this option to yes. Valid options
299 are: yes, no and ask.
300 protocol prot-id
302 On the TCP/IP link the specified protocol is used. Valid protocol
303 specifiers are A, B, and C.
304 Protocol A: write IO is reported as completed, if it has reached
306 local disk and local TCP send buffer.
307 Protocol B: write IO is reported as completed, if it has reached
309 local disk and remote buffer cache.
310 Protocol C: write IO is reported as completed, if it has reached
312 both local and remote disk.
313 device name minor nr
315 The name of the block device node of the resource being described.
317 You must use this device with your application (file system) and
318 you must not use the low level block device which is specified with
319 the disk parameter.
320 One can ether omit the name or minor and the minor number. If you
322 omit the name a default of /dev/drbdminor will be used.
323 Udev will create additional symlinks in /dev/drbd/by-res and
325 /dev/drbd/by-disk.
326 disk name
328 DRBD uses this block device to actually store and retrieve the
330 data. Never access such a device while DRBD is running on top of
331 it. This also holds true for dumpe2fs(8) and similar commands.
332 address AF addr:port
334 A resource needs one IP address per device, which is used to wait
336 for incoming connections from the partner device respectively to
337 reach the partner device. AF must be one of ipv4, ipv6, ssocks or
338 sdp (for compatibility reasons sci is an alias for ssocks). It may
339 be omited for IPv4 addresses. The actual IPv6 address that follows
340 the ipv6 keyword must be placed inside brackets: ipv6
341 [fd01:2345:6789:abcd::1]:7800.
342 Each DRBD resource needs a TCP port which is used to connect to the
344 node's partner device. Two different DRBD resources may not use the
345 same addr:port combination on the same node.
346 meta-disk internal,
348 meta-disk device,
349 meta-disk device [index]
350 Internal means that the last part of the backing device is used to
352 store the meta-data. The size of the meta-data is computed based on
353 the size of the device.
354 When a device is specified, either with or without an index, DRBD
356 stores the meta-data on this device. Without index, the size of the
357 meta-data is determined by the size of the data device. This is
358 usually used with LVM, which allows to have many variable sized
359 block devices. The meta-data size is 36kB + Backing-Storage-size /
360 32k, rounded up to the next 4kb boundary. (Rule of the thumb:
361 32kByte per 1GByte of storage, rounded up to the next MB.)
362 When an index is specified, each index number refers to a fixed
364 slot of meta-data of 128 MB, which allows a maximum data size of 4
365 TiB. This way, multiple DBRD devices can share the same meta-data
366 device. For example, if /dev/sde6[0] and /dev/sde6[1] are used,
367 /dev/sde6 must be at least 256 MB big. Because of the hard size
368 limit, use of meta-disk indexes is discouraged.
369 on-io-error handler
371 handler is taken, if the lower level device reports io-errors to
372 the upper layers.
373 handler may be pass_on, call-local-io-error or detach.
375 pass_on: The node downgrades the disk status to inconsistent, marks
377 the erroneous block as inconsistent in the bitmap and retries the
378 IO on the remote node.
379 call-local-io-error: Call the handler script local-io-error.
381 detach: The node drops its low level device, and continues in
383 diskless mode.
384 fencing fencing_policy
386 By fencing we understand preventive measures to avoid situations
388 where both nodes are primary and disconnected (AKA split brain).
389 Valid fencing policies are:
391 dont-care
393 This is the default policy. No fencing actions are taken.
394 resource-only
396 If a node becomes a disconnected primary, it tries to fence the
397 peer's disk. This is done by calling the fence-peer handler.
398 The handler is supposed to reach the other node over
399 alternative communication paths and call 'drbdadm outdate res'
400 there.
401 resource-and-stonith
403 If a node becomes a disconnected primary, it freezes all its IO
404 operations and calls its fence-peer handler. The fence-peer
405 handler is supposed to reach the peer over alternative
406 communication paths and call 'drbdadm outdate res' there. In
407 case it cannot reach the peer it should stonith the peer. IO is
408 resumed as soon as the situation is resolved. In case your
409 handler fails, you can resume IO with the resume-io command.
410 disk-barrier,
412 disk-flushes,
413 disk-drain
414 DRBD has four implementations to express write-after-write
415 dependencies to its backing storage device. DRBD will use the first
416 method that is supported by the backing storage device and that is
417 not disabled. By default the flush method is used.
418 Since drbd-8.4.2 disk-barrier is disabled by default because since
420 linux-2.6.36 (or 2.6.32 RHEL6) there is no reliable way to
421 determine if queuing of IO-barriers works. Dangerous only enable
422 if you are told so by one that knows for sure.
423 When selecting the method you should not only base your decision on
425 the measurable performance. In case your backing storage device has
426 a volatile write cache (plain disks, RAID of plain disks) you
427 should use one of the first two. In case your backing storage
428 device has battery-backed write cache you may go with option 3.
429 Option 4 (disable everything, use "none") is dangerous on most IO
430 stacks, may result in write-reordering, and if so, can
431 theoretically be the reason for data corruption, or disturb the
432 DRBD protocol, causing spurious disconnect/reconnect cycles. Do
433 not use no-disk-drain.
434 Unfortunately device mapper (LVM) might not support barriers.
436 The letter after "wo:" in /proc/drbd indicates with method is
438 currently in use for a device: b, f, d, n. The implementations are:
439 barrier
441 The first requires that the driver of the backing storage
442 device support barriers (called 'tagged command queuing' in
443 SCSI and 'native command queuing' in SATA speak). The use of
444 this method can be enabled by setting the disk-barrier options
445 to yes.
446 flush
448 The second requires that the backing device support disk
449 flushes (called 'force unit access' in the drive vendors
450 speak). The use of this method can be disabled setting
451 disk-flushes to no.
452 drain
454 The third method is simply to let write requests drain before
455 write requests of a new reordering domain are issued. This was
456 the only implementation before 8.0.9.
457 none
459 The fourth method is to not express write-after-write
460 dependencies to the backing store at all, by also specifying
461 no-disk-drain. This is dangerous on most IO stacks, may result
462 in write-reordering, and if so, can theoretically be the reason
463 for data corruption, or disturb the DRBD protocol, causing
464 spurious disconnect/reconnect cycles. Do not use
465 no-disk-drain.
466 md-flushes
468 Disables the use of disk flushes and barrier BIOs when accessing
469 the meta data device. See the notes on disk-flushes.
470 max-bio-bvecs
472 In some special circumstances the device mapper stack manages to
473 pass BIOs to DRBD that violate the constraints that are set forth
474 by DRBD's merge_bvec() function and which have more than one bvec.
475 A known example is: phys-disk -> DRBD -> LVM -> Xen -> misaligned
476 partition (63) -> DomU FS. Then you might see "bio would need to,
477 but cannot, be split:" in the Dom0's kernel log.
478 The best workaround is to proper align the partition within the VM
480 (E.g. start it at sector 1024). This costs 480 KiB of storage.
481 Unfortunately the default of most Linux partitioning tools is to
482 start the first partition at an odd number (63). Therefore most
483 distribution's install helpers for virtual linux machines will end
484 up with misaligned partitions. The second best workaround is to
485 limit DRBD's max bvecs per BIO (= max-bio-bvecs) to 1, but that
486 might cost performance.
487 The default value of max-bio-bvecs is 0, which means that there is
489 no user imposed limitation.
490 disk-timeout
492 If the lower-level device on which a DRBD device stores its data
493 does not finish an I/O request within the defined disk-timeout,
494 DRBD treats this as a failure. The lower-level device is detached,
495 and the device's disk state advances to Diskless. If DRBD is
496 connected to one or more peers, the failed request is passed on to
497 one of them.
498 This option is dangerous and may lead to kernel panic!
500 "Aborting" requests, or force-detaching the disk, is intended for
502 completely blocked/hung local backing devices which do no longer
503 complete requests at all, not even do error completions. In this
504 situation, usually a hard-reset and failover is the only way out.
505 By "aborting", basically faking a local error-completion, we allow
507 for a more graceful swichover by cleanly migrating services. Still
508 the affected node has to be rebooted "soon".
509 By completing these requests, we allow the upper layers to re-use
511 the associated data pages.
512 If later the local backing device "recovers", and now DMAs some
514 data from disk into the original request pages, in the best case it
515 will just put random data into unused pages; but typically it will
516 corrupt meanwhile completely unrelated data, causing all sorts of
517 damage.
518 Which means delayed successful completion, especially for READ
520 requests, is a reason to panic(). We assume that a delayed *error*
521 completion is OK, though we still will complain noisily about it.
522 The default value of disk-timeout is 0, which stands for an
524 infinite timeout. Timeouts are specified in units of 0.1 seconds.
525 This option is available since DRBD 8.3.12.
526 discard-zeroes-if-aligned {yes | no}
528 There are several aspects to discard/trim/unmap support on linux
530 block devices. Even if discard is supported in general, it may fail
531 silently, or may partially ignore discard requests. Devices also
532 announce whether reading from unmapped blocks returns defined data
533 (usually zeroes), or undefined data (possibly old data, possibly
534 garbage).
535 If on different nodes, DRBD is backed by devices with differing
537 discard characteristics, discards may lead to data divergence (old
538 data or garbage left over on one backend, zeroes due to unmapped
539 areas on the other backend). Online verify would now potentially
540 report tons of spurious differences. While probably harmless for
541 most use cases (fstrim on a file system), DRBD cannot have that.
542 To play safe, we have to disable discard support, if our local
544 backend (on a Primary) does not support "discard_zeroes_data=true".
545 We also have to translate discards to explicit zero-out on the
546 receiving side, unless the receiving side (Secondary) supports
547 "discard_zeroes_data=true", thereby allocating areas what were
548 supposed to be unmapped.
549 There are some devices (notably the LVM/DM thin provisioning) that
551 are capable of discard, but announce discard_zeroes_data=false. In
552 the case of DM-thin, discards aligned to the chunk size will be
553 unmapped, and reading from unmapped sectors will return zeroes.
554 However, unaligned partial head or tail areas of discard requests
555 will be silently ignored.
556 If we now add a helper to explicitly zero-out these unaligned
558 partial areas, while passing on the discard of the aligned full
559 chunks, we effectively achieve discard_zeroes_data=true on such
560 devices.
561 Setting discard-zeroes-if-aligned to yes will allow DRBD to use
563 discards, and to announce discard_zeroes_data=true, even on
564 backends that announce discard_zeroes_data=false.
565 Setting discard-zeroes-if-aligned to no will cause DRBD to always
567 fall-back to zero-out on the receiving side, and to not even
568 announce discard capabilities on the Primary, if the respective
569 backend announces discard_zeroes_data=false.
570 We used to ignore the discard_zeroes_data setting completely. To
572 not break established and expected behaviour, and suddenly cause
573 fstrim on thin-provisioned LVs to run out-of-space instead of
574 freeing up space, the default value is yes.
575 This option is available since 8.4.7.
577 --disable-write-same {yes | no}
579 Some disks announce WRITE_SAME support to the kernel but fail with
581 an I/O error upon actually receiving such a request. This mostly
582 happens when using virtualized disks -- notably, this behavior has
583 been observed with VMware's virtual disks.
584 When disable-write-same is set to yes, WRITE_SAME detection is
586 manually overriden and support is disabled.
587 The default value of disable-write-same is no. This option is
589 available since 8.4.7.
590 read-balancing method
592 The supported methods for load balancing of read requests are
593 prefer-local, prefer-remote, round-robin, least-pending,
594 when-congested-remote, 32K-striping, 64K-striping, 128K-striping,
595 256K-striping, 512K-striping and 1M-striping.
596 The default value of read-balancing is prefer-local. This option is
598 available since 8.4.1.
599 rs-discard-granularity byte
601 When rs-discard-granularity is set to a non zero, positive value
602 then DRBD tries to do a resync operation in requests of this size.
603 In case such a block contains only zero bytes on the sync source
604 node, the sync target node will issue a discard/trim/unmap command
605 for the area.
606 The value is constrained by the discard granularity of the backing
608 block device. In case rs-discard-granularity is not a multiplier of
609 the discard granularity of the backing block device DRBD rounds it
610 up. The feature only gets active if the backing block device reads
611 back zeroes after a discard command.
612 The default value of rs-discard-granularity is 0. This option is
614 available since 8.4.7.
615 sndbuf-size size
617 size is the size of the TCP socket send buffer. The default value
618 is 0, i.e. autotune. You can specify smaller or larger values.
619 Larger values are appropriate for reasonable write throughput with
620 protocol A over high latency networks. Values below 32K do not make
621 sense. Since 8.0.13 resp. 8.2.7, setting the size value to 0 means
622 that the kernel should autotune this.
623 rcvbuf-size size
625 size is the size of the TCP socket receive buffer. The default
626 value is 0, i.e. autotune. You can specify smaller or larger
627 values. Usually this should be left at its default. Setting the
628 size value to 0 means that the kernel should autotune this.
629 timeout time
631 If the partner node fails to send an expected response packet
633 within time tenths of a second, the partner node is considered dead
634 and therefore the TCP/IP connection is abandoned. This must be
635 lower than connect-int and ping-int. The default value is 60 = 6
636 seconds, the unit 0.1 seconds.
637 connect-int time
639 In case it is not possible to connect to the remote DRBD device
641 immediately, DRBD keeps on trying to connect. With this option you
642 can set the time between two retries. The default value is 10
643 seconds, the unit is 1 second.
644 ping-int time
646 If the TCP/IP connection linking a DRBD device pair is idle for
648 more than time seconds, DRBD will generate a keep-alive packet to
649 check if its partner is still alive. The default is 10 seconds, the
650 unit is 1 second.
651 ping-timeout time
653 The time the peer has time to answer to a keep-alive packet. In
655 case the peer's reply is not received within this time period, it
656 is considered as dead. The default value is 500ms, the default unit
657 are tenths of a second.
658 max-buffers number
660 Limits the memory usage per DRBD minor device on the receiving
662 side, or for internal buffers during resync or online-verify. Unit
663 is PAGE_SIZE, which is 4 KiB on most systems. The minimum possible
664 setting is hard coded to 32 (=128 KiB). These buffers are used to
665 hold data blocks while they are written to/read from disk. To avoid
666 possible distributed deadlocks on congestion, this setting is used
667 as a throttle threshold rather than a hard limit. Once more than
668 max-buffers pages are in use, further allocation from this pool is
669 throttled. You want to increase max-buffers if you cannot saturate
670 the IO backend on the receiving side.
671 ko-count number
673 In case the secondary node fails to complete a single write request
675 for count times the timeout, it is expelled from the cluster. (I.e.
676 the primary node will kill and restart the connection.) To disable
677 this feature, you should explicitly set it to 0; defaults may
678 change between versions.
679 max-epoch-size number
681 The highest number of data blocks between two write barriers. If
683 you set this smaller than 10, you might decrease your performance.
684 allow-two-primaries
686 With this option set you may assign the primary role to both nodes.
688 You only should use this option if you use a shared storage file
689 system on top of DRBD. At the time of writing the only ones are:
690 OCFS2 and GFS. If you use this option with any other file system,
691 you are going to crash your nodes and to corrupt your data!
692 unplug-watermark number
694 This setting has no effect with recent kernels that use explicit
695 on-stack plugging (upstream Linux kernel 2.6.39, distributions may
696 have backported).
697 When the number of pending write requests on the standby
699 (secondary) node exceeds the unplug-watermark, we trigger the
700 request processing of our backing storage device. Some storage
701 controllers deliver better performance with small values, others
702 deliver best performance when the value is set to the same value as
703 max-buffers, yet others don't feel much effect at all. Minimum 16,
704 default 128, maximum 131072.
705 cram-hmac-alg
707 You need to specify the HMAC algorithm to enable peer
709 authentication at all. You are strongly encouraged to use peer
710 authentication. The HMAC algorithm will be used for the challenge
711 response authentication of the peer. You may specify any digest
712 algorithm that is named in /proc/crypto.
713 shared-secret
715 The shared secret used in peer authentication. May be up to 64
717 characters. Note that peer authentication is disabled as long as no
718 cram-hmac-alg (see above) is specified.
719 after-sb-0pri policy
721 possible policies are:
722 disconnect
724 No automatic resynchronization, simply disconnect.
725 discard-younger-primary
727 Auto sync from the node that was primary before the split-brain
728 situation happened.
729 discard-older-primary
731 Auto sync from the node that became primary as second during
732 the split-brain situation.
733 discard-zero-changes
735 In case one node did not write anything since the split brain
736 became evident, sync from the node that wrote something to the
737 node that did not write anything. In case none wrote anything
738 this policy uses a random decision to perform a "resync" of 0
739 blocks. In case both have written something this policy
740 disconnects the nodes.
741 discard-least-changes
743 Auto sync from the node that touched more blocks during the
744 split brain situation.
745 discard-node-NODENAME
747 Auto sync to the named node.
748 after-sb-1pri policy
750 possible policies are:
751 disconnect
753 No automatic resynchronization, simply disconnect.
754 consensus
756 Discard the version of the secondary if the outcome of the
757 after-sb-0pri algorithm would also destroy the current
758 secondary's data. Otherwise disconnect.
759 violently-as0p
761 Always take the decision of the after-sb-0pri algorithm, even
762 if that causes an erratic change of the primary's view of the
763 data. This is only useful if you use a one-node FS (i.e. not
764 OCFS2 or GFS) with the allow-two-primaries flag, AND if you
765 really know what you are doing. This is DANGEROUS and MAY CRASH
766 YOUR MACHINE if you have an FS mounted on the primary node.
767 discard-secondary
769 Discard the secondary's version.
770 call-pri-lost-after-sb
772 Always honor the outcome of the after-sb-0pri algorithm. In
773 case it decides the current secondary has the right data, it
774 calls the "pri-lost-after-sb" handler on the current primary.
775 after-sb-2pri policy
777 possible policies are:
778 disconnect
780 No automatic resynchronization, simply disconnect.
781 violently-as0p
783 Always take the decision of the after-sb-0pri algorithm, even
784 if that causes an erratic change of the primary's view of the
785 data. This is only useful if you use a one-node FS (i.e. not
786 OCFS2 or GFS) with the allow-two-primaries flag, AND if you
787 really know what you are doing. This is DANGEROUS and MAY CRASH
788 YOUR MACHINE if you have an FS mounted on the primary node.
789 call-pri-lost-after-sb
791 Call the "pri-lost-after-sb" helper program on one of the
792 machines. This program is expected to reboot the machine, i.e.
793 make it secondary.
794 always-asbp
796 Normally the automatic after-split-brain policies are only used if
797 current states of the UUIDs do not indicate the presence of a third
798 node.
799 With this option you request that the automatic after-split-brain
801 policies are used as long as the data sets of the nodes are somehow
802 related. This might cause a full sync, if the UUIDs indicate the
803 presence of a third node. (Or double faults led to strange UUID
804 sets.)
805 rr-conflict policy
807 This option helps to solve the cases when the outcome of the resync
808 decision is incompatible with the current role assignment in the
809 cluster.
810 disconnect
812 No automatic resynchronization, simply disconnect.
813 violently
815 Sync to the primary node is allowed, violating the assumption
816 that data on a block device are stable for one of the nodes.
817 Dangerous, do not use.
818 call-pri-lost
820 Call the pri-lost-after-sb helper program on one of the
821 machines unless that machine can demote to secondary. The
822 helper program is expected to reboot the machine, which brings
823 the node into a secondary role. Which machine runs the helper
824 program is determined by the after-sb-0pri strategy.
825 data-integrity-alg alg
827 DRBD can ensure the data integrity of the user's data on the
828 network by comparing hash values. Normally this is ensured by the
829 16 bit checksums in the headers of TCP/IP packets.
830 This option can be set to any of the kernel's data digest
832 algorithms. In a typical kernel configuration you should have at
833 least one of md5, sha1, and crc32c available. By default this is
834 not enabled.
835 See also the notes on data integrity.
837 tcp-cork
839 DRBD usually uses the TCP socket option TCP_CORK to hint to the
840 network stack when it can expect more data, and when it should
841 flush out what it has in its send queue. It turned out that there
842 is at least one network stack that performs worse when one uses
843 this hinting method. Therefore we introducted this option. By
844 setting tcp-cork to no you can disable the setting and clearing of
845 the TCP_CORK socket option by DRBD.
846 on-congestion congestion_policy,
848 congestion-fill fill_threshold,
849 congestion-extents active_extents_threshold
850 By default DRBD blocks when the available TCP send queue becomes
851 full. That means it will slow down the application that generates
852 the write requests that cause DRBD to send more data down that TCP
853 connection.
854 When DRBD is deployed with DRBD-proxy it might be more desirable
856 that DRBD goes into AHEAD/BEHIND mode shortly before the send queue
857 becomes full. In AHEAD/BEHIND mode DRBD does no longer replicate
858 data, but still keeps the connection open.
859 The advantage of the AHEAD/BEHIND mode is that the application is
861 not slowed down, even if DRBD-proxy's buffer is not sufficient to
862 buffer all write requests. The downside is that the peer node falls
863 behind, and that a resync will be necessary to bring it back into
864 sync. During that resync the peer node will have an inconsistent
865 disk.
866 Available congestion_policys are block and pull-ahead. The default
868 is block. Fill_threshold might be in the range of 0 to 10GiBytes.
869 The default is 0 which disables the check.
870 Active_extents_threshold has the same limits as al-extents.
871 The AHEAD/BEHIND mode and its settings are available since DRBD
873 8.3.10.
874 wfc-timeout time
876 Wait for connection timeout.
877 The init script drbd(8) blocks the boot process until the DRBD
879 resources are connected. When the cluster manager starts later, it
880 does not see a resource with internal split-brain. In case you want
881 to limit the wait time, do it here. Default is 0, which means
882 unlimited. The unit is seconds.
883 degr-wfc-timeout time
885 Wait for connection timeout, if this node was a degraded cluster.
887 In case a degraded cluster (= cluster with only one node left) is
888 rebooted, this timeout value is used instead of wfc-timeout,
889 because the peer is less likely to show up in time, if it had been
890 dead before. Value 0 means unlimited.
891 outdated-wfc-timeout time
893 Wait for connection timeout, if the peer was outdated. In case a
895 degraded cluster (= cluster with only one node left) with an
896 outdated peer disk is rebooted, this timeout value is used instead
897 of wfc-timeout, because the peer is not allowed to become primary
898 in the meantime. Value 0 means unlimited.
899 wait-after-sb
901 By setting this option you can make the init script to continue to
902 wait even if the device pair had a split brain situation and
903 therefore refuses to connect.
904 become-primary-on node-name
906 Sets on which node the device should be promoted to primary role by
907 the init script. The node-name might either be a host name or the
908 keyword both. When this option is not set the devices stay in
909 secondary role on both nodes. Usually one delegates the role
910 assignment to a cluster manager (e.g. heartbeat).
911 stacked-timeouts
913 Usually wfc-timeout and degr-wfc-timeout are ignored for stacked
914 devices, instead twice the amount of connect-int is used for the
915 connection timeouts. With the stacked-timeouts keyword you disable
916 this, and force DRBD to mind the wfc-timeout and degr-wfc-timeout
917 statements. Only do that if the peer of the stacked resource is
918 usually not available or will usually not become primary. By using
919 this option incorrectly, you run the risk of causing unexpected
920 split brain.
921 resync-rate rate
923 To ensure a smooth operation of the application on top of DRBD, it
925 is possible to limit the bandwidth which may be used by background
926 synchronizations. The default is 250 KB/sec, the default unit is
927 KB/sec. Optional suffixes K, M, G are allowed.
928 use-rle
930 During resync-handshake, the dirty-bitmaps of the nodes are
932 exchanged and merged (using bit-or), so the nodes will have the
933 same understanding of which blocks are dirty. On large devices, the
934 fine grained dirty-bitmap can become large as well, and the bitmap
935 exchange can take quite some time on low-bandwidth links.
936 Because the bitmap typically contains compact areas where all bits
938 are unset (clean) or set (dirty), a simple run-length encoding
939 scheme can considerably reduce the network traffic necessary for
940 the bitmap exchange.
941 For backward compatibility reasons, and because on fast links this
943 possibly does not improve transfer time but consumes cpu cycles,
944 this defaults to off.
945 socket-check-timeout value
947 In setups involving a DRBD-proxy and connections that experience a
949 lot of buffer-bloat it might be necessary to set ping-timeout to an
950 unusual high value. By default DRBD uses the same value to wait if
951 a newly established TCP-connection is stable. Since the DRBD-proxy
952 is usually located in the same data center such a long wait time
953 may hinder DRBD's connect process.
954 In such setups socket-check-timeout should be set to at least to
956 the round trip time between DRBD and DRBD-proxy. I.e. in most cases
957 to 1.
958 The default unit is tenths of a second, the default value is 0
960 (which causes DRBD to use the value of ping-timeout instead).
961 Introduced in 8.4.5.
962 resync-after res-name
964 By default, resynchronization of all devices would run in parallel.
966 By defining a resync-after dependency, the resynchronization of
967 this resource will start only if the resource res-name is already
968 in connected state (i.e., has finished its resynchronization).
969 al-extents extents
971 DRBD automatically performs hot area detection. With this parameter
973 you control how big the hot area (= active set) can get. Each
974 extent marks 4M of the backing storage (= low-level device). In
975 case a primary node leaves the cluster unexpectedly, the areas
976 covered by the active set must be resynced upon rejoining of the
977 failed node. The data structure is stored in the meta-data area,
978 therefore each change of the active set is a write operation to the
979 meta-data device. A higher number of extents gives longer resync
980 times but less updates to the meta-data. The default number of
981 extents is 1237. (Minimum: 7, Maximum: 65534)
982 Note that the effective maximum may be smaller, depending on how
984 you created the device meta data, see also drbdmeta(8). The
985 effective maximum is 919 * (available on-disk activity-log
986 ring-buffer area/4kB -1), the default 32kB ring-buffer effects a
987 maximum of 6433 (covers more than 25 GiB of data). We recommend to
988 keep this well within the amount your backend storage and
989 replication link are able to resync inside of about 5 minutes.
990 al-updates {yes | no}
992 DRBD's activity log transaction writing makes it possible, that
994 after the crash of a primary node a partial (bit-map based) resync
995 is sufficient to bring the node back to up-to-date. Setting
996 al-updates to no might increase normal operation performance but
997 causes DRBD to do a full resync when a crashed primary gets
998 reconnected. The default value is yes.
999 verify-alg hash-alg
1001 During online verification (as initiated by the verify
1002 sub-command), rather than doing a bit-wise comparison, DRBD applies
1003 a hash function to the contents of every block being verified, and
1004 compares that hash with the peer. This option defines the hash
1005 algorithm being used for that purpose. It can be set to any of the
1006 kernel's data digest algorithms. In a typical kernel configuration
1007 you should have at least one of md5, sha1, and crc32c available. By
1008 default this is not enabled; you must set this option explicitly in
1009 order to be able to use on-line device verification.
1010 See also the notes on data integrity.
1012 csums-alg hash-alg
1014 A resync process sends all marked data blocks from the source to
1015 the destination node, as long as no csums-alg is given. When one is
1016 specified the resync process exchanges hash values of all marked
1017 blocks first, and sends only those data blocks that have different
1018 hash values.
1019 This setting is useful for DRBD setups with low bandwidth links.
1021 During the restart of a crashed primary node, all blocks covered by
1022 the activity log are marked for resync. But a large part of those
1023 will actually be still in sync, therefore using csums-alg will
1024 lower the required bandwidth in exchange for CPU cycles.
1025 c-plan-ahead plan_time,
1027 c-fill-target fill_target,
1028 c-delay-target delay_target,
1029 c-max-rate max_rate
1030 The dynamic resync speed controller gets enabled with setting
1031 plan_time to a positive value. It aims to fill the buffers along
1032 the data path with either a constant amount of data fill_target, or
1033 aims to have a constant delay time of delay_target along the path.
1034 The controller has an upper bound of max_rate.
1035 By plan_time the agility of the controller is configured. Higher
1037 values yield for slower/lower responses of the controller to
1038 deviation from the target value. It should be at least 5 times RTT.
1039 For regular data paths a fill_target in the area of 4k to 100k is
1040 appropriate. For a setup that contains drbd-proxy it is advisable
1041 to use delay_target instead. Only when fill_target is set to 0 the
1042 controller will use delay_target. 5 times RTT is a reasonable
1043 starting value. Max_rate should be set to the bandwidth available
1044 between the DRBD-hosts and the machines hosting DRBD-proxy, or to
1045 the available disk-bandwidth.
1046 The default value of plan_time is 0, the default unit is 0.1
1048 seconds. Fill_target has 0 and sectors as default unit.
1049 Delay_target has 1 (100ms) and 0.1 as default unit. Max_rate has
1050 10240 (100MiB/s) and KiB/s as default unit.
1051 The dynamic resync speed controller and its settings are available
1053 since DRBD 8.3.9.
1054 c-min-rate min_rate
1056 A node that is primary and sync-source has to schedule application
1057 IO requests and resync IO requests. The min_rate tells DRBD use
1058 only up to min_rate for resync IO and to dedicate all other
1059 available IO bandwidth to application requests.
1060 Note: The value 0 has a special meaning. It disables the limitation
1062 of resync IO completely, which might slow down application IO
1063 considerably. Set it to a value of 1, if you prefer that resync IO
1064 never slows down application IO.
1065 Note: Although the name might suggest that it is a lower bound for
1067 the dynamic resync speed controller, it is not. If the DRBD-proxy
1068 buffer is full, the dynamic resync speed controller is free to
1069 lower the resync speed down to 0, completely independent of the
1070 c-min-rate setting.
1071 The default value of min_rate is 250, in units of KiB/s
1073 on-no-data-accessible ond-policy
1075 This setting controls what happens to IO requests on a degraded,
1076 disk less node (I.e. no data store is reachable). The available
1077 policies are io-error and suspend-io.
1078 If ond-policy is set to suspend-io you can either resume IO by
1080 attaching/connecting the last lost data storage, or by the drbdadm
1081 resume-io res command. The latter will result in IO errors of
1082 course.
1083 The default is io-error. This setting is available since DRBD
1085 8.3.9.
1086 cpu-mask cpu-mask
1088 Sets the cpu-affinity-mask for DRBD's kernel threads of this
1090 device. The default value of cpu-mask is 0, which means that DRBD's
1091 kernel threads should be spread over all CPUs of the machine. This
1092 value must be given in hexadecimal notation. If it is too big it
1093 will be truncated.
1094 pri-on-incon-degr cmd
1096 This handler is called if the node is primary, degraded and if the
1098 local copy of the data is inconsistent.
1099 pri-lost-after-sb cmd
1101 The node is currently primary, but lost the after-split-brain auto
1103 recovery procedure. As as consequence, it should be abandoned.
1104 pri-lost cmd
1106 The node is currently primary, but DRBD's algorithm thinks that it
1108 should become sync target. As a consequence it should give up its
1109 primary role.
1110 fence-peer cmd
1112 The handler is part of the fencing mechanism. This handler is
1114 called in case the node needs to fence the peer's disk. It should
1115 use other communication paths than DRBD's network link.
1116 local-io-error cmd
1118 DRBD got an IO error from the local IO subsystem.
1120 initial-split-brain cmd
1122 DRBD has connected and detected a split brain situation. This
1124 handler can alert someone in all cases of split brain, not just
1125 those that go unresolved.
1126 split-brain cmd
1128 DRBD detected a split brain situation but remains unresolved.
1130 Manual recovery is necessary. This handler should alert someone on
1131 duty.
1132 before-resync-target cmd
1134 DRBD calls this handler just before a resync begins on the node
1136 that becomes resync target. It might be used to take a snapshot of
1137 the backing block device.
1138 after-resync-target cmd
1140 DRBD calls this handler just after a resync operation finished on
1142 the node whose disk just became consistent after being inconsistent
1143 for the duration of the resync. It might be used to remove a
1144 snapshot of the backing device that was created by the
1145 before-resync-target handler.
1146 Other Keywords
1148 include file-pattern
1149 Include all files matching the wildcard pattern file-pattern. The
1151 include statement is only allowed on the top level, i.e. it is not
1152 allowed inside any section.
1153
1155 There are two independent methods in DRBD to ensure the integrity of
1156 the mirrored data. The online-verify mechanism and the
1157 data-integrity-alg of the network section.
1158 Both mechanisms might deliver false positives if the user of DRBD
1160 modifies the data which gets written to disk while the transfer goes
1161 on. This may happen for swap, or for certain append while global sync,
1162 or truncate/rewrite workloads, and not necessarily poses a problem for
1163 the integrity of the data. Usually when the initiator of the data
1164 transfer does this, it already knows that that data block will not be
1165 part of an on disk data structure, or will be resubmitted with correct
1166 data soon enough.
1167 The data-integrity-alg causes the receiving side to log an error about
1169 "Digest integrity check FAILED: Ns +x\n", where N is the sector offset,
1170 and x is the size of the request in bytes. It will then disconnect, and
1171 reconnect, thus causing a quick resync. If the sending side at the same
1172 time detected a modification, it warns about "Digest mismatch, buffer
1173 modified by upper layers during write: Ns +x\n", which shows that this
1174 was a false positive. The sending side may detect these buffer
1175 modifications immediately after the unmodified data has been copied to
1176 the tcp buffers, in which case the receiving side won't notice it.
1177 The most recent (2007) example of systematic corruption was an issue
1179 with the TCP offloading engine and the driver of a certain type of GBit
1180 NIC. The actual corruption happened on the DMA transfer from core
1181 memory to the card. Since the TCP checksum gets calculated on the card,
1182 this type of corruption stays undetected as long as you do not use
1183 either the online verify or the data-integrity-alg.
1184 We suggest to use the data-integrity-alg only during a pre-production
1186 phase due to its CPU costs. Further we suggest to do online verify runs
1187 regularly e.g. once a month during a low load period.
1188
1190 This document was revised for version 8.4.0 of the DRBD distribution.
1191
1193 Written by Philipp Reisner <philipp.reisner@linbit.com> and Lars
1194 Ellenberg <lars.ellenberg@linbit.com>.
1195
1197 Report bugs to <drbd-user@lists.linbit.com>.
1198
1200 Copyright 2001-2008 LINBIT Information Technologies, Philipp Reisner,
1201 Lars Ellenberg. This is free software; see the source for copying
1202 conditions. There is NO warranty; not even for MERCHANTABILITY or
1203 FITNESS FOR A PARTICULAR PURPOSE.
1204
1206 drbd(8), drbddisk(8), drbdsetup(8), drbdmeta(8), drbdadm(8), DRBD
1207 User's Guide[1], DRBD web site[3]
1208
1210 1. DRBD User's Guide
1211 http://www.drbd.org/users-guide/
1212 2. DRBD's online usage counter
1214 http://usage.drbd.org
1215 3. DRBD web site
1217 http://www.drbd.org/
1218DRBD 8.4.0 6 May 2011 DRBD.CONF(5)