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 In this example, there is a single DRBD resource (called r0) which uses
18 protocol C for the connection between its devices. The device which
19 runs on host alice uses /dev/drbd1 as devices for its application, and
20 /dev/sda7 as low-level storage for the data. The IP addresses are used
21 to specify the networking interfaces to be used. An eventually running
22 resync process should use about 10MByte/second of IO bandwidth.
23 There may be multiple resource sections in a single drbd.conf file. For
25 more examples, please have a look at the DRBD User's Guide[1].
26
28 The file consists of sections and parameters. A section begins with a
29 keyword, sometimes an additional name, and an opening brace (“{”). A
30 section ends with a closing brace (“}”. The braces enclose the
31 parameters.
32 section [name] { parameter value; [...] }
34 A parameter starts with the identifier of the parameter followed by
36 whitespace. Every subsequent character is considered as part of the
37 parameter's value. A special case are Boolean parameters which consist
38 only of the identifier. Parameters are terminated by a semicolon (“;”).
39 Some parameter values have default units which might be overruled by K,
41 M or G. These units are defined in the usual way (K = 2^10 = 1024, M =
42 1024 K, G = 1024 M).
43 Comments may be placed into the configuration file and must begin with
45 a hash sign (“#”). Subsequent characters are ignored until the end of
46 the line.
47 Sections
49 skip
50 Comments out chunks of text, even spanning more than one line.
51 Characters between the keyword skip and the opening brace (“{”) are
52 ignored. Everything enclosed by the braces is skipped. This comes
53 in handy, if you just want to comment out some 'resource [name]
54 {...}' section: just precede it with '“skip”'.
55 global
57 Configures some global parameters. Currently only minor-count,
58 dialog-refresh, disable-ip-verification and usage-count are allowed
59 here. You may only have one global section, preferably as the first
60 section.
61 common
63 All resources inherit the options set in this section. The common
64 section might have a startup, a syncer, a handlers, a net and a
65 disk section.
66 resource name
68 Configures a DRBD resource. Each resource section needs to have two
69 (or more) on host sections and may have a startup, a syncer, a
70 handlers, a net and a disk section. Required parameter in this
71 section: protocol.
72 on host-name
74 Carries the necessary configuration parameters for a DRBD device of
75 the enclosing resource. host-name is mandatory and must match the
76 Linux host name (uname -n) of one of the nodes. You may list more
77 than one host name here, in case you want to use the same
78 parameters on several hosts (you'd have to move the IP around
79 usually). Or you may list more than two such sections.
80 resource r1 {
82 protocol C;
83 device minor 1;
84 meta-disk internal;
85 on alice bob {
87 address 10.2.2.100:7801;
88 disk /dev/mapper/some-san;
89 }
90 on charlie {
91 address 10.2.2.101:7801;
92 disk /dev/mapper/other-san;
93 }
94 on daisy {
95 address 10.2.2.103:7801;
96 disk /dev/mapper/other-san-as-seen-from-daisy;
97 }
98 }
99 See also the floating section keyword. Required parameters in this
102 section: device, disk, address, meta-disk, flexible-meta-disk.
103 stacked-on-top-of resource
105 For a stacked DRBD setup (3 or 4 nodes), a stacked-on-top-of is
106 used instead of an on section. Required parameters in this section:
107 device and address.
108 floating AF addr:port
110 Carries the necessary configuration parameters for a DRBD device of
111 the enclosing resource. This section is very similar to the on
112 section. The difference to the on section is that the matching of
113 the host sections to machines is done by the IP-address instead of
114 the node name. Required parameters in this section: device, disk,
115 meta-disk, flexible-meta-disk, all of which may be inherited from
116 the resource section, in which case you may shorten this section
117 down to just the address identifier.
118 resource r2 {
120 protocol C;
121 device minor 2;
122 disk /dev/sda7;
123 meta-disk internal;
124 # short form, device, disk and meta-disk inherited
126 floating 10.1.1.31:7802;
127 # longer form, only device inherited
129 floating 10.1.1.32:7802 {
130 disk /dev/sdb;
131 meta-disk /dev/sdc8;
132 }
133 }
134 disk
138 This section is used to fine tune DRBD's properties in respect to
139 the low level storage. Please refer to drbdsetup(8) for detailed
140 description of the parameters. Optional parameters: on-io-error,
141 size, fencing, use-bmbv, no-disk-barrier, no-disk-flushes,
142 no-disk-drain, no-md-flushes, max-bio-bvecs.
143 net
145 This section is used to fine tune DRBD's properties. Please refer
146 to drbdsetup(8) for a detailed description of this section's
147 parameters. Optional parameters: sndbuf-size, rcvbuf-size, timeout,
148 connect-int, ping-int, ping-timeout, max-buffers, max-epoch-size,
149 ko-count, allow-two-primaries, cram-hmac-alg, shared-secret,
150 after-sb-0pri, after-sb-1pri, after-sb-2pri, data-integrity-alg,
151 no-tcp-cork
152 startup
154 This section is used to fine tune DRBD's properties. Please refer
155 to drbdsetup(8) for a detailed description of this section's
156 parameters. Optional parameters: wfc-timeout, degr-wfc-timeout,
157 outdated-wfc-timeout, wait-after-sb, stacked-timeouts and
158 become-primary-on.
159 syncer
161 This section is used to fine tune the synchronization daemon for
162 the device. Please refer to drbdsetup(8) for a detailed description
163 of this section's parameters. Optional parameters: rate, after,
164 al-extents, use-rle, cpu-mask, verify-alg, csums-alg,
165 delay-probe-volume, delay-probe-interval, throttle-threshold and
166 hold-off-threshold.
167 handlers
169 In this section you can define handlers (executables) that are
170 started by the DRBD system in response to certain events. Optional
171 parameters: pri-on-incon-degr, pri-lost-after-sb, pri-lost,
172 fence-peer (formerly oudate-peer), local-io-error,
173 initial-split-brain, split-brain, before-resync-target,
174 after-resync-target.
175 The interface is done via environment variables:
177 DRBD_PEER is deprecated.
179 Please note that not all of these might be set for all handlers,
181 and that some values might not be useable for a floating
182 definition.
183 Parameters
185 minor-count count
186 may be a number from 1 to 255.
187 Use minor-count if you want to define massively more resources
189 later without reloading the DRBD kernel module. Per default the
190 module loads with 11 more resources than you have currently in your
191 config but at least 32.
192 dialog-refresh time
194 may be 0 or a positive number.
195 The user dialog redraws the second count every time seconds (or
197 does no redraws if time is 0). The default value is 1.
198 disable-ip-verification
200 Use disable-ip-verification if, for some obscure reasons, drbdadm
201 can/might not use ip or ifconfig to do a sanity check for the IP
202 address. You can disable the IP verification with this option.
203 usage-count val
205 Please participate in DRBD's online usage counter[2]. The most
206 convenient way to do so is to set this option to yes. Valid options
207 are: yes, no and ask.
208 protocol prot-id
210 On the TCP/IP link the specified protocol is used. Valid protocol
211 specifiers are A, B, and C.
212 Protocol A: write IO is reported as completed, if it has reached
214 local disk and local TCP send buffer.
215 Protocol B: write IO is reported as completed, if it has reached
217 local disk and remote buffer cache.
218 Protocol C: write IO is reported as completed, if it has reached
220 both local and remote disk.
221 device name minor nr
223 The name of the block device node of the resource being described.
224 You must use this device with your application (file system) and
225 you must not use the low level block device which is specified with
226 the disk parameter.
227 One can ether omit the name or minor and the minor number. If you
229 omit the name a default of /dev/drbdminor will be used.
230 Udev will create additional symlinks in /dev/drbd/by-res and
232 /dev/drbd/by-disk.
233 disk name
235 DRBD uses this block device to actually store and retrieve the
236 data. Never access such a device while DRBD is running on top of
237 it. This also holds true for dumpe2fs(8) and similar commands.
238 address AF addr:port
240 A resource needs one IP address per device, which is used to wait
241 for incoming connections from the partner device respectively to
242 reach the partner device. AF must be one of ipv4, ipv6, ssocks or
243 sdp (for compatibility reasons sci is an alias for ssocks). It may
244 be omited for IPv4 addresses. The actual IPv6 address that follows
245 the ipv6 keyword must be placed inside brackets: ipv6
246 [fd01:2345:6789:abcd::1]:7800.
247 Each DRBD resource needs a TCP port which is used to connect to the
249 node's partner device. Two different DRBD resources may not use the
250 same addr:port combination on the same node.
251 meta-disk internal, flexible-meta-disk internal, meta-disk device
253 [index], flexible-meta-disk device
254 Internal means that the last part of the backing device is used to
255 store the meta-data. You must not use [index] with internal. Note:
256 Regardless of whether you use the meta-disk or the
257 flexible-meta-disk keyword, it will always be of the size needed
258 for the remaining storage size.
259 You can use a single block device to store meta-data of multiple
261 DRBD devices. E.g. use meta-disk /dev/sde6[0]; and meta-disk
262 /dev/sde6[1]; for two different resources. In this case the
263 meta-disk would need to be at least 256 MB in size.
264 With the flexible-meta-disk keyword you specify a block device as
266 meta-data storage. You usually use this with LVM, which allows you
267 to have many variable sized block devices. The required size of the
268 meta-disk block device is 36kB + Backing-Storage-size / 32k. Round
269 this number to the next 4kb boundary up and you have the exact
270 size. Rule of the thumb: 32kByte per 1GByte of storage, round up to
271 the next MB.
272 on-io-error handler
274 is taken, if the lower level device reports io-errors to the upper
275 layers.
276 handler may be pass_on, call-local-io-error or detach.
278 pass_on: Report the io-error to the upper layers. On Primary report
280 it to the mounted file system. On Secondary ignore it.
281 call-local-io-error: Call the handler script local-io-error.
283 detach: The node drops its low level device, and continues in
285 diskless mode.
286 fencing fencing_policy
288 By fencing we understand preventive measures to avoid situations
289 where both nodes are primary and disconnected (AKA split brain).
290 Valid fencing policies are:
292 dont-care
294 This is the default policy. No fencing actions are taken.
295 resource-only
297 If a node becomes a disconnected primary, it tries to fence the
298 peer's disk. This is done by calling the fence-peer handler.
299 The handler is supposed to reach the other node over
300 alternative communication paths and call 'drbdadm outdate res'
301 there.
302 resource-and-stonith
304 If a node becomes a disconnected primary, it freezes all its IO
305 operations and calls its fence-peer handler. The fence-peer
306 handler is supposed to reach the peer over alternative
307 communication paths and call 'drbdadm outdate res' there. In
308 case it cannot reach the peer it should stonith the peer. IO is
309 resumed as soon as the situation is resolved. In case your
310 handler fails, you can resume IO with the resume-io command.
311 use-bmbv
313 In case the backing storage's driver has a merge_bvec_fn()
314 function, DRBD has to pretend that it can only process IO requests
315 in units not larger than 4KiB. (At the time of writing the only
316 known drivers which have such a function are: md (software raid
317 driver), dm (device mapper - LVM) and DRBD itself).
318 To get the best performance out of DRBD on top of software RAID (or
320 any other driver with a merge_bvec_fn() function) you might enable
321 this function, if you know for sure that the merge_bvec_fn()
322 function will deliver the same results on all nodes of your
323 cluster. I.e. the physical disks of the software RAID are of
324 exactly the same type. Use this option only if you know what you
325 are doing.
326 no-disk-barrier, no-disk-flushes, no-disk-drain
328 DRBD has four implementations to express write-after-write
329 dependencies to its backing storage device. DRBD will use the first
330 method that is supported by the backing storage device and that is
331 not disabled by the user.
332 When selecting the method you should not only base your decision on
334 the measurable performance. In case your backing storage device has
335 a volatile write cache (plain disks, RAID of plain disks) you
336 should use one of the first two. In case your backing storage
337 device has battery-backed write cache you may go with option 3 or
338 4. Option 4 will deliver the best performance on such devices.
339 Unfortunately device mapper (LVM) might not support barriers.
341 The letter after "wo:" in /proc/drbd indicates with method is
343 currently in use for a device: b, f, d, n. The implementations are:
344 barrier
346 The first requires that the driver of the backing storage
347 device support barriers (called 'tagged command queuing' in
348 SCSI and 'native command queuing' in SATA speak). The use of
349 this method can be disabled by the no-disk-barrier option.
350 flush
352 The second requires that the backing device support disk
353 flushes (called 'force unit access' in the drive vendors
354 speak). The use of this method can be disabled using the
355 no-disk-flushes option.
356 drain
358 The third method is simply to let write requests drain before
359 write requests of a new reordering domain are issued. This was
360 the only implementation before 8.0.9. You can disable this
361 method by using the no-disk-drain option.
362 none
364 The fourth method is to not express write-after-write
365 dependencies to the backing store at all.
366 no-md-flushes
368 Disables the use of disk flushes and barrier BIOs when accessing
369 the meta data device. See the notes on no-disk-flushes.
370 max-bio-bvecs
372 In some special circumstances the device mapper stack manages to
373 pass BIOs to DRBD that violate the constraints that are set forth
374 by DRBD's merge_bvec() function and which have more than one bvec.
375 A known example is: phys-disk -> DRBD -> LVM -> Xen -> misaligned
376 partition (63) -> DomU FS. Then you might see "bio would need to,
377 but cannot, be split:" in the Dom0's kernel log.
378 The best workaround is to proper align the partition within the VM
380 (E.g. start it at sector 1024). This costs 480 KiB of storage.
381 Unfortunately the default of most Linux partitioning tools is to
382 start the first partition at an odd number (63). Therefore most
383 distribution's install helpers for virtual linux machines will end
384 up with misaligned partitions. The second best workaround is to
385 limit DRBD's max bvecs per BIO (= max-bio-bvecs) to 1, but that
386 might cost performance.
387 The default value of max-bio-bvecs is 0, which means that there is
389 no user imposed limitation.
390 sndbuf-size size
392 is the size of the TCP socket send buffer. The default value is 0,
393 i.e. autotune. You can specify smaller or larger values. Larger
394 values are appropriate for reasonable write throughput with
395 protocol A over high latency networks. Values below 32K do not make
396 sense. Since 8.0.13 resp. 8.2.7, setting the size value to 0 means
397 that the kernel should autotune this.
398 rcvbuf-size size
400 is the size of the TCP socket receive buffer. The default value is
401 0, i.e. autotune. You can specify smaller or larger values. Usually
402 this should be left at its default. Setting the size value to 0
403 means that the kernel should autotune this.
404 timeout time
406 If the partner node fails to send an expected response packet
407 within time tenths of a second, the partner node is considered dead
408 and therefore the TCP/IP connection is abandoned. This must be
409 lower than connect-int and ping-int. The default value is 60 = 6
410 seconds, the unit 0.1 seconds.
411 connect-int time
413 In case it is not possible to connect to the remote DRBD device
414 immediately, DRBD keeps on trying to connect. With this option you
415 can set the time between two retries. The default value is 10
416 seconds, the unit is 1 second.
417 ping-int time
419 If the TCP/IP connection linking a DRBD device pair is idle for
420 more than time seconds, DRBD will generate a keep-alive packet to
421 check if its partner is still alive. The default is 10 seconds, the
422 unit is 1 second.
423 ping-timeout time
425 The time the peer has time to answer to a keep-alive packet. In
426 case the peer's reply is not received within this time period, it
427 is considered as dead. The default value is 500ms, the default unit
428 are tenths of a second.
429 max-buffers number
431 Maximum number of requests to be allocated by DRBD. Unit is
432 PAGE_SIZE, which is 4 KiB on most systems. The minimum is hard
433 coded to 32 (=128 KiB). For high-performance installations it might
434 help if you increase that number. These buffers are used to hold
435 data blocks while they are written to disk.
436 ko-count number
438 In case the secondary node fails to complete a single write request
439 for count times the timeout, it is expelled from the cluster. (I.e.
440 the primary node goes into StandAlone mode.) The default value is
441 0, which disables this feature.
442 max-epoch-size number
444 The highest number of data blocks between two write barriers. If
445 you set this smaller than 10, you might decrease your performance.
446 allow-two-primaries
448 With this option set you may assign the primary role to both nodes.
449 You only should use this option if you use a shared storage file
450 system on top of DRBD. At the time of writing the only ones are:
451 OCFS2 and GFS. If you use this option with any other file system,
452 you are going to crash your nodes and to corrupt your data!
453 unplug-watermark number
455 When the number of pending write requests on the standby
456 (secondary) node exceeds the unplug-watermark, we trigger the
457 request processing of our backing storage device. Some storage
458 controllers deliver better performance with small values, others
459 deliver best performance when the value is set to the same value as
460 max-buffers. Minimum 16, default 128, maximum 131072.
461 cram-hmac-alg
463 You need to specify the HMAC algorithm to enable peer
464 authentication at all. You are strongly encouraged to use peer
465 authentication. The HMAC algorithm will be used for the challenge
466 response authentication of the peer. You may specify any digest
467 algorithm that is named in /proc/crypto.
468 shared-secret
470 The shared secret used in peer authentication. May be up to 64
471 characters. Note that peer authentication is disabled as long as no
472 cram-hmac-alg (see above) is specified.
473 after-sb-0pri policy
475 possible policies are:
476 disconnect
478 No automatic resynchronization, simply disconnect.
479 discard-younger-primary
481 Auto sync from the node that was primary before the split-brain
482 situation happened.
483 discard-older-primary
485 Auto sync from the node that became primary as second during
486 the split-brain situation.
487 discard-zero-changes
489 In case one node did not write anything since the split brain
490 became evident, sync from the node that wrote something to the
491 node that did not write anything. In case none wrote anything
492 this policy uses a random decision to perform a "resync" of 0
493 blocks. In case both have written something this policy
494 disconnects the nodes.
495 discard-least-changes
497 Auto sync from the node that touched more blocks during the
498 split brain situation.
499 discard-node-NODENAME
501 Auto sync to the named node.
502 after-sb-1pri policy
504 possible policies are:
505 disconnect
507 No automatic resynchronization, simply disconnect.
508 consensus
510 Discard the version of the secondary if the outcome of the
511 after-sb-0pri algorithm would also destroy the current
512 secondary's data. Otherwise disconnect.
513 violently-as0p
515 Always take the decision of the after-sb-0pri algorithm, even
516 if that causes an erratic change of the primary's view of the
517 data. This is only useful if you use a one-node FS (i.e. not
518 OCFS2 or GFS) with the allow-two-primaries flag, AND if you
519 really know what you are doing. This is DANGEROUS and MAY CRASH
520 YOUR MACHINE if you have an FS mounted on the primary node.
521 discard-secondary
523 Discard the secondary's version.
524 call-pri-lost-after-sb
526 Always honor the outcome of the after-sb-0pri algorithm. In
527 case it decides the current secondary has the right data, it
528 calls the "pri-lost-after-sb" handler on the current primary.
529 after-sb-2pri policy
531 possible policies are:
532 disconnect
534 No automatic resynchronization, simply disconnect.
535 violently-as0p
537 Always take the decision of the after-sb-0pri algorithm, even
538 if that causes an erratic change of the primary's view of the
539 data. This is only useful if you use a one-node FS (i.e. not
540 OCFS2 or GFS) with the allow-two-primaries flag, AND if you
541 really know what you are doing. This is DANGEROUS and MAY CRASH
542 YOUR MACHINE if you have an FS mounted on the primary node.
543 call-pri-lost-after-sb
545 Call the "pri-lost-after-sb" helper program on one of the
546 machines. This program is expected to reboot the machine, i.e.
547 make it secondary.
548 always-asbp
550 Normally the automatic after-split-brain policies are only used if
551 current states of the UUIDs do not indicate the presence of a third
552 node.
553 With this option you request that the automatic after-split-brain
555 policies are used as long as the data sets of the nodes are somehow
556 related. This might cause a full sync, if the UUIDs indicate the
557 presence of a third node. (Or double faults led to strange UUID
558 sets.)
559 rr-conflict policy
561 This option helps to solve the cases when the outcome of the resync
562 decision is incompatible with the current role assignment in the
563 cluster.
564 disconnect
566 No automatic resynchronization, simply disconnect.
567 violently
569 Sync to the primary node is allowed, violating the assumption
570 that data on a block device are stable for one of the nodes.
571 Dangerous, do not use.
572 call-pri-lost
574 Call the "pri-lost" helper program on one of the machines. This
575 program is expected to reboot the machine, i.e. make it
576 secondary.
577 data-integrity-alg alg
579 DRBD can ensure the data integrity of the user's data on the
580 network by comparing hash values. Normally this is ensured by the
581 16 bit checksums in the headers of TCP/IP packets.
582 This option can be set to any of the kernel's data digest
584 algorithms. In a typical kernel configuration you should have at
585 least one of md5, sha1, and crc32c available. By default this is
586 not enabled.
587 See also the notes on data integrity.
589 no-tcp-cork
591 DRBD usually uses the TCP socket option TCP_CORK to hint to the
592 network stack when it can expect more data, and when it should
593 flush out what it has in its send queue. It turned out that there
594 is at least one network stack that performs worse when one uses
595 this hinting method. Therefore we introducted this option, which
596 disables the setting and clearing of the TCP_CORK socket option by
597 DRBD.
598 wfc-timeout time
600 Wait for connection timeout. The init script drbd(8) blocks the
601 boot process until the DRBD resources are connected. When the
602 cluster manager starts later, it does not see a resource with
603 internal split-brain. In case you want to limit the wait time, do
604 it here. Default is 0, which means unlimited. The unit is seconds.
605 degr-wfc-timeout time
607 Wait for connection timeout, if this node was a degraded cluster.
608 In case a degraded cluster (= cluster with only one node left) is
609 rebooted, this timeout value is used instead of wfc-timeout,
610 because the peer is less likely to show up in time, if it had been
611 dead before. Value 0 means unlimited.
612 outdated-wfc-timeout time
614 Wait for connection timeout, if the peer was outdated. In case a
615 degraded cluster (= cluster with only one node left) with an
616 outdated peer disk is rebooted, this timeout value is used instead
617 of wfc-timeout, because the peer is not allowed to become primary
618 in the meantime. Value 0 means unlimited.
619 wait-after-sb
621 By setting this option you can make the init script to continue to
622 wait even if the device pair had a split brain situation and
623 therefore refuses to connect.
624 become-primary-on node-name
626 Sets on which node the device should be promoted to primary role by
627 the init script. The node-name might either be a host name or the
628 keyword both. When this option is not set the devices stay in
629 secondary role on both nodes. Usually one delegates the role
630 assignment to a cluster manager (e.g. heartbeat).
631 stacked-timeouts
633 Usually wfc-timeout and degr-wfc-timeout are ignored for stacked
634 devices, instead twice the amount of connect-int is used for the
635 connection timeouts. With the stacked-timeouts keyword you disable
636 this, and force DRBD to mind the wfc-timeout and degr-wfc-timeout
637 statements. Only do that if the peer of the stacked resource is
638 usually not available or will usually not become primary. By using
639 this option incorrectly, you run the risk of causing unexpected
640 split brain.
641 rate rate
643 To ensure a smooth operation of the application on top of DRBD, it
644 is possible to limit the bandwidth which may be used by background
645 synchronizations. The default is 250 KB/sec, the default unit is
646 KB/sec. Optional suffixes K, M, G are allowed.
647 use-rle
649 During resync-handshake, the dirty-bitmaps of the nodes are
650 exchanged and merged (using bit-or), so the nodes will have the
651 same understanding of which blocks are dirty. On large devices, the
652 fine grained dirty-bitmap can become large as well, and the bitmap
653 exchange can take quite some time on low-bandwidth links.
654 Because the bitmap typically contains compact areas where all bits
656 are unset (clean) or set (dirty), a simple run-length encoding
657 scheme can considerably reduce the network traffic necessary for
658 the bitmap exchange.
659 For backward compatibilty reasons, and because on fast links this
661 possibly does not improve transfer time but consumes cpu cycles,
662 this defaults to off.
663 after res-name
665 By default, resynchronization of all devices would run in parallel.
666 By defining a sync-after dependency, the resynchronization of this
667 resource will start only if the resource res-name is already in
668 connected state (i.e., has finished its resynchronization).
669 al-extents extents
671 DRBD automatically performs hot area detection. With this parameter
672 you control how big the hot area (= active set) can get. Each
673 extent marks 4M of the backing storage (= low-level device). In
674 case a primary node leaves the cluster unexpectedly, the areas
675 covered by the active set must be resynced upon rejoining of the
676 failed node. The data structure is stored in the meta-data area,
677 therefore each change of the active set is a write operation to the
678 meta-data device. A higher number of extents gives longer resync
679 times but less updates to the meta-data. The default number of
680 extents is 127. (Minimum: 7, Maximum: 3843)
681 verify-alg hash-alg
683 During online verification (as initiated by the verify
684 sub-command), rather than doing a bit-wise comparison, DRBD applies
685 a hash function to the contents of every block being verified, and
686 compares that hash with the peer. This option defines the hash
687 algorithm being used for that purpose. It can be set to any of the
688 kernel's data digest algorithms. In a typical kernel configuration
689 you should have at least one of md5, sha1, and crc32c available. By
690 default this is not enabled; you must set this option explicitly in
691 order to be able to use on-line device verification.
692 See also the notes on data integrity.
694 csums-alg hash-alg
696 A resync process sends all marked data blocks from the source to
697 the destination node, as long as no csums-alg is given. When one is
698 specified the resync process exchanges hash values of all marked
699 blocks first, and sends only those data blocks that have different
700 hash values.
701 This setting is useful for DRBD setups with low bandwidth links.
703 During the restart of a crashed primary node, all blocks covered by
704 the activity log are marked for resync. But a large part of those
705 will actually be still in sync, therefore using csums-alg will
706 lower the required bandwidth in exchange for CPU cycles.
707 delay-probe-volume bytes, delay-probe-interval interval,
709 throttle-threshold throttle_delay, hold-off-threshold hold_off_delay
710 During resync at least every bytes of data and at least every
711 interval * 100ms a pair of delay probes get inserted in DRBD's
712 packet stream. Those packets are used to measure the delay of
713 packts on the data socket caused by queuing in various network
714 components along the path.
715 If the delay on the data socket becomes greater than throttle_delay
717 DRBD will slow down the resync in order to keep the delay small.
718 The resync speed gets linearly slowed down it reaches 0 at a delay
719 of hold_off_delay.
720 cpu-mask cpu-mask
722 Sets the cpu-affinity-mask for DRBD's kernel threads of this
723 device. The default value of cpu-mask is 0, which means that DRBD's
724 kernel threads should be spread over all CPUs of the machine. This
725 value must be given in hexadecimal notation. If it is too big it
726 will be truncated.
727 pri-on-incon-degr cmd
729 This handler is called if the node is primary, degraded and if the
730 local copy of the data is inconsistent.
731 pri-lost-after-sb cmd
733 The node is currently primary, but lost the after-split-brain auto
734 recovery procedure. As as consequence, it should be abandoned.
735 pri-lost cmd
737 The node is currently primary, but DRBD's algorithm thinks that it
738 should become sync target. As a consequence it should give up its
739 primary role.
740 fence-peer cmd
742 The handler is part of the fencing mechanism. This handler is
743 called in case the node needs to fence the peer's disk. It should
744 use other communication paths than DRBD's network link.
745 local-io-error cmd
747 DRBD got an IO error from the local IO subsystem.
748 initial-split-brain cmd
750 DRBD has connected and detected a split brain situation. This
751 handler can alert someone in all cases of split brain, not just
752 those that go unresolved.
753 split-brain cmd
755 DRBD detected a split brain situation but remains unresolved.
756 Manual recovery is necessary. This handler should alert someone on
757 duty.
758 before-resync-target cmd
760 DRBD calls this handler just before a resync begins on the node
761 that becomes resync target. It might be used to take a snapshot of
762 the backing block device.
763 after-resync-target cmd
765 DRBD calls this handler just after a resync operation finished on
766 the node whose disk just became consistent after being inconsistent
767 for the duration of the resync. It might be used to remove a
768 snapshot of the backing device that was created by the
769 before-resync-target handler.
770 Other Keywords
772 include file-pattern
773 Include all files matching the wildcard pattern file-pattern. The
774 include statement is only allowed on the top level, i.e. it is not
775 allowed inside any section.
776
778 There are two independent methods in DRBD to ensure the integrity of
779 the mirrored data. The online-verify mechanism and the
780 data-integrity-alg of the network section.
781 Both mechanisms might deliver false positives if the user of DRBD
783 modifies the data which gets written to disk while the transfer goes
784 on. Currently the swap code and ReiserFS are known to do so. In both
785 cases this is not a problem, because when the initiator of the data
786 transfer does this it already knows that that data block will not be
787 part of an on disk data structure.
788 The most recent (2007) example of systematically corruption was an
790 issue with the TCP offloading engine and the driver of a certain type
791 of GBit NIC. The actual corruption happened on the DMA transfer from
792 core memory to the card. Since the TCP checksum gets calculated on the
793 card this type of corruption stays undetected as long as you do not use
794 either the online verify or the data-integrity-alg.
795 We suggest to use the data-integrity-alg only during a pre-production
797 phase due to its CPU costs. Further we suggest to do online verify runs
798 regularly e.g. once a month during a low load period.
799
801 This document was revised for version 8.3.2 of the DRBD distribution.
802
804 Written by Philipp Reisner philipp.reisner@linbit.com and Lars
805 Ellenberg lars.ellenberg@linbit.com.
806
808 Report bugs to drbd-user@lists.linbit.com.
809
811 Copyright 2001-2008 LINBIT Information Technologies, Philipp Reisner,
812 Lars Ellenberg. This is free software; see the source for copying
813 conditions. There is NO warranty; not even for MERCHANTABILITY or
814 FITNESS FOR A PARTICULAR PURPOSE.
815
817 drbd(8), drbddisk(8), drbdsetup(8), drbdadm(8), DRBD User's Guide[1],
818 DRBD web site[3]
819
821 1. DRBD User's Guide
822 http://www.drbd.org/users-guide/
823 2. DRBD's online usage counter
825 http://usage.drbd.org
826 3. DRBD web site
828 http://www.drbd.org/
829DRBD 8.3.2 5 Dec 2008 DRBD.CONF(5)