1 commit 586538e22afba85c59beda49789ec42024e7a061
2 Author: Gary Lowell <gary.lowell@inktank.com>
3 Date: Tue Jan 29 23:54:47 2013 -0800
7 commit bcb8dfad9cbb4c6af7ae7f9584e36449a03cd1b6
8 Author: Dan Mick <dan.mick@inktank.com>
9 Date: Tue Jan 29 23:05:49 2013 -0800
11 cls_rbd, cls_rgw: use PRI*64 when printing/logging 64-bit values
13 caused segfaults in 32-bit build
16 Signed-off-by: Dan Mick <dan.mick@inktank.com>
17 Reviewed-by: Sage Weil <sage@inktank.com>
18 (cherry picked from commit e253830abac76af03c63239302691f7fac1af381)
20 commit 5a7c5088cc8f57f75eb594a21bf5fb6661e50978
21 Author: Dan Mick <dan.mick@inktank.com>
22 Date: Tue Jan 29 15:18:53 2013 -0800
24 init-ceph: make ulimit -n be part of daemon command
26 ulimit -n from 'max open files' was being set only on the machine
27 running /etc/init.d/ceph. It needs to be added to the commands to
28 start the daemons, and run both locally and remotely.
30 Verified by examining /proc/<pid>/limits on local and remote hosts
33 Signed-off-by: Dan Mick <dan.mick@inktank.com>
34 Reviewed-by: Loïc Dachary <loic@dachary.org>
35 Reviewed-by: Gary Lowell <gary.lowell@inktank.com>
36 (cherry picked from commit 84a024b647c0ac2ee5a91bacdd4b8c966e44175c)
38 commit 95677fc599b9bf37ab4c2037b3675fd68f92ebcf
39 Author: Joao Eduardo Luis <joao.luis@inktank.com>
40 Date: Sat Jan 12 01:06:36 2013 +0000
42 mon: OSDMonitor: only share osdmap with up OSDs
44 Try to share the map with a randomly picked OSD; if the picked monitor is
45 not 'up', then try to find the nearest 'up' OSD in the map by doing a
46 backward and a forward linear search on the map -- this would be O(n) in
47 the worst case scenario, as we only do a single iteration starting on the
48 picked position, incrementing and decrementing two different iterators
49 until we find an appropriate OSD or we exhaust the map.
54 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
55 Reviewed-by: Sage Weil <sage@inktank.com>
56 (cherry picked from commit 3610e72e4f9117af712f34a2e12c5e9537a5746f)
58 commit e4d76cb8594c0ec901f89c2f2e8cc53e00eb2a06
59 Author: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
60 Date: Sun Jan 27 21:57:31 2013 +0100
62 utime: fix narrowing conversion compiler warning in sleep()
65 ./include/utime.h: In member function 'void utime_t::sleep()':
66 ./include/utime.h:139:50: warning: narrowing conversion of
67 '((utime_t*)this)->utime_t::tv.utime_t::<anonymous struct>::tv_sec' from
68 '__u32 {aka unsigned int}' to '__time_t {aka long int}' inside { } is
69 ill-formed in C++11 [-Wnarrowing]
70 ./include/utime.h:139:50: warning: narrowing conversion of
71 '((utime_t*)this)->utime_t::tv.utime_t::<anonymous struct>::tv_nsec' from
72 '__u32 {aka unsigned int}' to 'long int' inside { } is
73 ill-formed in C++11 [-Wnarrowing]
75 Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
76 (cherry picked from commit 014fc6d6c1c68e2e3ad0117d08c4e46e4030d49e)
78 commit a8964107ddf02ac4a6707a997e1b634c1084a3b9
79 Author: Yehuda Sadeh <yehuda@inktank.com>
80 Date: Mon Jan 28 17:13:23 2013 -0800
82 rgw: fix crash when missing content-type in POST object
85 This fixes a crash when handling S3 POST request and content type
88 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
89 (cherry picked from commit f41010c44b3a4489525d25cd35084a168dc5f537)
91 commit 11e1f3acf0953e9ac38322c0423144eaabd7bb61
92 Author: Samuel Just <sam.just@inktank.com>
93 Date: Fri Jan 11 15:00:02 2013 -0800
95 ReplicatedPG: make_snap_collection when moving snap link in snap_trimmer
98 Signed-off-by: Samuel Just <sam.just@inktank.com>
99 Reviewed-by: Sage Weil <sage@inktank.com>
100 (cherry picked from commit 88956e3186798058a1170803f8abfc0f3cf77a07)
102 commit c9201d0e9de5f4766a2d9f4715eb7c69691964de
103 Author: Samuel Just <sam.just@inktank.com>
104 Date: Fri Jan 11 16:43:14 2013 -0800
106 ReplicatedPG: correctly handle new snap collections on replica
109 Signed-off-by: Samuel Just <sam.just@inktank.com>
110 Reviewed-by: Sage Weil <sage@inktank.com>
111 (cherry picked from commit 9e44fca13bf1ba39dbcad29111b29f46c49d59f7)
113 commit 2efdfb41c1bc9128b76416630ee00a75de90c020
114 Author: Joao Eduardo Luis <joao.luis@inktank.com>
115 Date: Sun Jan 27 18:08:15 2013 +0000
117 mon: Elector: reset the acked leader when the election finishes and we lost
119 Failure to do so will mean that we will always ack the same leader during
120 an election started by another monitor. This had been working so far
121 because we were still acking the existing leader if he was supposed to
122 still be the leader; or we were acking a new potentially leader; or we
123 would eventually fall behind on an election and start a new election
124 ourselves, thus resetting the previously acked leader. While this wasn't
125 something that mattered much until now, the timechecks code stumbled into
126 this tiny issue and was failing hard at completing a round because there
127 wouldn't be a reset before the election started -- timechecks are bound
132 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
133 (cherry picked from commit c54781618569680898e77e151dd7364f22ac4aa1)
135 commit a16c6f3dc278e19e66776ffde45de3ff0db46a6c
136 Author: Josh Durgin <josh.durgin@inktank.com>
137 Date: Wed Dec 26 14:24:22 2012 -0800
139 rbd: fix bench-write infinite loop
141 I/O was continously submitted as long as there were few enough ops in
142 flight. If the number of 'threads' was high, or caching was turned on,
143 there would never be that many ops in flight, so the loop would continue
144 indefinitely. Instead, submit at most io_threads ops per offset.
147 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
148 Reviewed-by: Dan Mick <dan.mick@inktank.com>
149 Reviewed-by: Sage Weil <sage.weil@inktank.com>
150 (cherry picked from commit d81ac8418f9e6bbc9adcc69b2e7cb98dd4db6abb)
152 commit 76f93751d3603e3fb5c4b9e14bfdac406d8d1a58
153 Author: Dan Mick <dan.mick@inktank.com>
154 Date: Fri Jan 4 18:00:24 2013 -0800
156 rbd: Don't call ProgressContext's finish() if there's an error.
158 do_copy was different from the others; call pc.fail() on error and
159 do not call pc.finish().
162 Signed-off-by: Dan Mick <dan.mick@inktank.com>
163 (cherry picked from commit 0978dc4963fe441fb67afecb074bc7b01798d59d)
165 commit 10053b14623f9c19727cb4d2d3a6b62945bef5c1
166 Author: Josh Durgin <josh.durgin@inktank.com>
167 Date: Wed Jan 2 14:15:24 2013 -0800
169 librbd: establish watch before reading header
171 This eliminates a window in which a race could occur when we have an
172 image open but no watch established. The previous fix (using
173 assert_version) did not work well with resend operations.
175 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
176 (cherry picked from commit c4370ff03f8ab655a009cfd9ba3a0827d8c58b11)
178 commit f666c617f6a5f8d94ce81461942c9f94a0775fb2
179 Author: Josh Durgin <josh.durgin@inktank.com>
180 Date: Wed Jan 2 12:32:33 2013 -0800
182 Revert "librbd: ensure header is up to date after initial read"
184 Using assert version for linger ops doesn't work with retries,
185 since the version will change after the first send.
186 This reverts commit e1776809031c6dad441cfb2b9fac9612720b9083.
190 qa/workunits/rbd/watch_correct_version.sh
191 (cherry picked from commit e0858fa89903cf4055889c405f17515504e917a0)
193 commit 575a58666adbca83d15468899272e8c369e903e1
194 Author: Sage Weil <sage@inktank.com>
195 Date: Wed Jan 23 22:16:49 2013 -0800
197 os/FileStore: only adjust up op queue for btrfs
199 We only need to adjust up the op queue limits during commit for btrfs,
200 because the snapshot initiation (async create) is currently
201 high-latency and the op queue is quiesced during that period.
203 This lets us revert 44dca5c, which disabled the extra allowance because
204 it is generally bad for non-btrfs writeahead mode.
206 Signed-off-by: Sage Weil <sage@inktank.com>
207 (cherry picked from commit 38871e27eca5a34de78db23aa3663f6cb045d461)
209 commit c9eb1b0a99b0e55f7d7343176dad17d1a53589a1
210 Author: Sage Weil <sage@inktank.com>
211 Date: Thu Jan 24 10:52:46 2013 -0800
213 common/HeartbeatMap: fix uninitialized variable
215 Introduced by me in 132045ce085e8584a3e177af552ee7a5205b13d8. Thank you,
218 Signed-off-by: Sage Weil <sage@inktank.com>
219 (cherry picked from commit 00cfe1d3af286ffab7660933415684f18449720c)
221 commit e6bceeedb0b77d23416560bd951326587470aacb
222 Author: Samuel Just <sam.just@inktank.com>
223 Date: Fri Jan 25 11:31:29 2013 -0800
225 sharedptr_registry: remove extaneous Mutex::Locker declaration
227 For some reason, the lookup() retry loop (for when happened to
228 race with a removal and grab an invalid WeakPtr) locked
229 the lock again. This causes the #3836 crash since the lock
230 is already locked. It's rare since it requires a lookup between
231 invalidation of the WeakPtr and removal of the WeakPtr entry.
235 Signed-off-by: Samuel Just <sam.just@inktank.com>
236 (cherry picked from commit 037900dc7a051ce2293a4ef9d0e71911b29ec159)
238 commit 60888cafdc53d6b381cd634170646c12669e1754
239 Author: Samuel Just <sam.just@inktank.com>
240 Date: Thu Jan 24 12:02:09 2013 -0800
242 FileStore: ping TPHandle after each operation in _do_transactions
244 Each completed operation in the transaction proves thread
245 liveness, a stuck thread should still trigger the timeouts.
249 Signed-off-by: Samuel Just <sam.just@inktank.com>
250 (cherry picked from commit 0c1cc687b6a40d3c6a26671f0652e1b51c3fd1af)
252 commit 6b8a673f88cbaca2891834dd5d2137a0e076fd1e
253 Author: Samuel Just <sam.just@inktank.com>
254 Date: Thu Jan 24 11:07:37 2013 -0800
256 OSD: use TPHandle in peering_wq
258 Implement _process overload with TPHandle argument and use
259 that to ping the hb map between pgs and between map epochs
260 when advancing a pg. The thread will still timeout if
261 genuinely stuck at any point.
265 Signed-off-by: Samuel Just <sam.just@inktank.com>
266 (cherry picked from commit e0511f4f4773766d04e845af2d079f82f3177cb6)
268 commit aa6d20aac22d4c14ff059dbc28e06b7a5e5d6de1
269 Author: Samuel Just <sam.just@inktank.com>
270 Date: Thu Jan 24 11:04:04 2013 -0800
272 WorkQueue: add TPHandle to allow _process to ping the hb map
275 Signed-off-by: Samuel Just <sam.just@inktank.com>
276 (cherry picked from commit 4f653d23999b24fc8c65a59f14905db6630be5b5)
278 commit e66a75052a340b15693f08b05f7f9f5d975b0978
279 Author: Samuel Just <sam.just@inktank.com>
280 Date: Wed Jan 23 12:49:04 2013 -0800
282 ReplicatedPG: handle omap > max_recovery_chunk
284 span_of fails if len == 0.
287 Signed-off-by: Samuel Just <sam.just@inktank.com>
288 (cherry picked from commit 8a97eef1f7004988449bd7ace4c69d5796495139)
290 commit 44f0407a6b259e87803539ec9e942043de0cf35d
291 Author: Samuel Just <sam.just@inktank.com>
292 Date: Wed Jan 23 12:18:31 2013 -0800
294 ReplicatedPG: correctly handle omap key larger than max chunk
297 Signed-off-by: Samuel Just <sam.just@inktank.com>
298 (cherry picked from commit c3dec3e30a85ecad0090c75a38f28cb83e36232e)
300 commit 50fd6ac9f147a4418d64dfe08843402e7cfb4910
301 Author: Samuel Just <sam.just@inktank.com>
302 Date: Wed Jan 23 12:15:10 2013 -0800
304 ReplicatedPG: start scanning omap at omap_recovered_to
306 Previously, we started scanning omap after omap_recovered_to.
307 This is a problem since the break in the loop implies that
308 omap_recovered_to is the first key not recovered.
311 Signed-off-by: Samuel Just <sam.just@inktank.com>
312 (cherry picked from commit 09c71f2f5ee9929ac4574f4c35fb8c0211aad097)
314 commit 4b32eecba2e2bd8e8ea17e1888e6971d31e71439
315 Author: Samuel Just <sam.just@inktank.com>
316 Date: Wed Jan 23 11:50:13 2013 -0800
318 ReplicatedPG: don't finish_recovery_op until the transaction completes
320 Signed-off-by: Samuel Just <sam.just@inktank.com>
321 (cherry picked from commit 62a4b96831c1726043699db86a664dc6a0af8637)
323 commit da34c77b93e3f880c01329711ab8eca7776b1830
324 Author: Samuel Just <sam.just@inktank.com>
325 Date: Wed Jan 23 11:35:47 2013 -0800
327 ReplicatedPG: ack push only after transaction has completed
329 Signed-off-by: Samuel Just <sam.just@inktank.com>
330 (cherry picked from commit 20278c4f77b890d5b2b95d2ccbeb4fbe106667ac)
332 commit f9381c74931b80294e5df60f6d2e69c946b8fe88
333 Author: Samuel Just <sam.just@inktank.com>
334 Date: Wed Jan 23 11:13:28 2013 -0800
336 ObjectStore: add queue_transactions with oncomplete
338 Signed-off-by: Samuel Just <sam.just@inktank.com>
339 (cherry picked from commit 4d6ba06309b80fb21de7bb5d12d5482e71de5f16)
341 commit e2560554f0568c30c786632723c5ce0c86043359
342 Author: Sage Weil <sage@inktank.com>
343 Date: Tue Jan 22 21:18:45 2013 -0800
345 common/HeartbeatMap: inject unhealthy heartbeat for N seconds
347 This lets us test code that is triggered by an unhealthy heartbeat in a
350 Signed-off-by: Sage Weil <sage@inktank.com>
351 (cherry picked from commit 132045ce085e8584a3e177af552ee7a5205b13d8)
353 commit cbe8b5bca40fd63a382b1a903087e7c34b314985
354 Author: Sage Weil <sage@inktank.com>
355 Date: Tue Jan 22 18:08:22 2013 -0800
357 os/FileStore: add stall injection into filestore op queue
359 Allow admin to artificially induce a stall in the op queue. Forces the
360 thread(s) to sleep for N seconds. We pause for 1 second increments and
361 recheck the value so that a previously stalled thread can be unwedged by
362 reinjecting a lower value (or 0). To stall indefinitely, just injust
365 Signed-off-by: Sage Weil <sage@inktank.com>
366 (cherry picked from commit 657df852e9c89bfacdbce25ea014f7830d61e6aa)
368 commit beb6ca44cd0e7fc405360e6da974252cb76e7039
369 Author: Sage Weil <sage@inktank.com>
370 Date: Tue Jan 22 18:03:10 2013 -0800
372 osd: do not join cluster if not healthy
374 If our internal heartbeats are failing, do not send a boot message and try
377 Signed-off-by: Sage Weil <sage@inktank.com>
378 (cherry picked from commit a4e78652cdd1698e8dd72dda51599348d013e5e0)
380 commit 1ecdfca3a3b4985ebd182a5f399c7b15af258663
381 Author: Sage Weil <sage@inktank.com>
382 Date: Tue Jan 22 18:01:07 2013 -0800
384 osd: hold lock while calling start_boot on startup
386 This probably doesn't strictly matter because start_boot doesn't need the
387 lock (currently) and few other threads should be running, but it is
388 better to be consistent.
390 Signed-off-by: Sage Weil <sage@inktank.com>
391 (cherry picked from commit c406476c0309792c43df512dddb2fe0f19835e71)
393 commit e120bf20b3c7213fbde20907e158792dd36c8e54
394 Author: Sage Weil <sage@inktank.com>
395 Date: Tue Jan 22 17:56:32 2013 -0800
397 osd: do not reply to ping if internal heartbeat is not healthy
399 If we find that our internal threads are stalled, do not reply to ping
400 requests. If we do this long enough, peers will mark us down. If we are
401 only transiently unhealthy, we will reply to the next ping and they will
402 be satisfied. If we are unhealthy and marked down, and eventually recover,
403 we will mark ourselves back up.
405 Signed-off-by: Sage Weil <sage@inktank.com>
406 (cherry picked from commit ad6b231127a6bfcbed600a7493ca3b66c68484d2)
408 commit 5f396e2b9360401dfe4dc2afa6acc37df8580c80
409 Author: Sage Weil <sage@inktank.com>
410 Date: Tue Jan 22 17:53:40 2013 -0800
412 osd: reduce op thread heartbeat default 30 -> 15 seconds
414 If the thread stalls for 15 seconds, let our internal heartbeat fail.
415 This will let us internally respond more quickly to a stalled or failing
418 Signed-off-by: Sage Weil <sage@inktank.com>
419 (cherry picked from commit 61eafffc3242357d9add48be9308222085536898)
421 commit fca288b718ef4582d65ff4b9d1fc87ba53d7fd8d
422 Author: Sage Weil <sage@inktank.com>
423 Date: Mon Jan 21 21:02:01 2013 -0800
425 osd: improve sub_op flag points
427 Signed-off-by: Sage Weil <sage@inktank.com>
428 (cherry picked from commit 73a969366c8bbd105579611320c43e2334907fef)
430 commit f13ddc8a2df401c37f6dc792eb93fc0cc45705e2
431 Author: Sage Weil <sage@inktank.com>
432 Date: Mon Jan 21 20:55:20 2013 -0800
434 osd: refactor ReplicatedPG::do_sub_op
436 PULL is the only case where we don't wait for active.
438 Signed-off-by: Sage Weil <sage@inktank.com>
439 (cherry picked from commit 23c02bce90c9725ccaf4295de3177e8146157723)
441 commit d5e00f963f177745f0e0684d5977460b7ab59fbd
442 Author: Sage Weil <sage@inktank.com>
443 Date: Mon Jan 21 16:36:36 2013 -0800
445 osd: make last state for slow requests more informative
447 Report on the last event string, and pass in important context for the
448 op event list, including:
450 - which peers were sent sub ops and we are waiting for
451 - which pg queue we are delayed by
453 Signed-off-by: Sage Weil <sage@inktank.com>
454 (cherry picked from commit a1137eb3e168c2d00f93789e4d565c1584790df0)
456 commit ab3a110cbe16b548bb96225656b64507aa67e78f
457 Author: Sage Weil <sage@inktank.com>
458 Date: Mon Jan 21 15:59:07 2013 -0800
460 osd: dump op priority queue state via admin socket
462 Signed-off-by: Sage Weil <sage@inktank.com>
463 (cherry picked from commit 24d0d7eb0165c8b8f923f2d8896b156bfb5e0e60)
465 commit 43a65d04d8a13621a856baec85fb741971c13cb0
466 Author: Sage Weil <sage@inktank.com>
467 Date: Mon Jan 21 15:50:33 2013 -0800
469 osd: simplify asok to single callback
471 Signed-off-by: Sage Weil <sage@inktank.com>
472 (cherry picked from commit 33efe32151e04beaafd9435d7f86dc2eb046214d)
474 commit d040798637da03e3df937181de156714fc62a550
475 Author: Sage Weil <sage@inktank.com>
476 Date: Mon Jan 21 15:58:57 2013 -0800
478 common/PrioritizedQueue: dump state to Formatter
480 Signed-off-by: Sage Weil <sage@inktank.com>
481 (cherry picked from commit 514af15e95604bd241d2a98a97b938889c6876db)
483 commit 691fd505ad606bd8befd2b19113ee51a17a0a543
484 Author: Sage Weil <sage@inktank.com>
485 Date: Mon Jan 21 15:29:28 2013 -0800
487 common/PrioritizedQueue: add min cost, max tokens per bucket
491 First, we need to cap the tokens per bucket. Otherwise, a stream of
492 items at one priority over time will indefinitely inflate the tokens
493 available at another priority. The cap should represent how "bursty"
494 we allow a given bucket to be. Start with 4MB for now.
496 Second, set a floor on the item cost. Otherwise, we can have an
497 infinite queue of 0 cost items that start over queues. More
498 realistically, we need to balance the overhead of processing small items
499 with the cost of large items. I.e., a 4 KB item is not 1/1000th as
500 expensive as a 4MB item.
502 Signed-off-by: Sage Weil <sage@inktank.com>
503 (cherry picked from commit 6e3363b20e590cd9df89f2caebe71867b94cc291)
505 commit a2b03fe08044b5c121ea6b4c2f9d19e73e4c83d1
506 Author: Sage Weil <sage@inktank.com>
507 Date: Mon Jan 21 14:52:54 2013 -0800
509 common/PrioritizedQueue: buckets -> tokens
511 Signed-off-by: Sage Weil <sage@inktank.com>
512 (cherry picked from commit c549a0cf6fae78c8418a3b4b0702fd8a1e4ce482)
514 commit 612d75cdee0daf9dfca97831c249e1ac3fbd59fc
515 Author: Sage Weil <sage@inktank.com>
516 Date: Mon Jan 21 14:31:00 2013 -0800
518 note puller's max chunk in pull requests
520 this lets us calculate a cost value
521 (cherry picked from commit 128fcfcac7d3fb66ca2c799df521591a98b82e05)
523 commit 2224e413fba11795693025fa8f11c3f1fba4bbaa
524 Author: Sage Weil <sage@inktank.com>
525 Date: Mon Jan 21 14:14:25 2013 -0800
527 osd: add OpRequest flag point when commit is sent
529 With writeahead journaling in particular, we can get requests that
530 stay in the queue for a long time even after the commit is sent to the
531 client while we are waiting for the transaction to apply to the fs.
532 Instead of showing up as 'waiting for subops', make it clear that the
533 client has gotten its reply and it is local state that is slow.
535 Signed-off-by: Sage Weil <sage@inktank.com>
536 (cherry picked from commit b685f727d4c37a26cb78bd4a04cce041428ceb52)
538 commit 5b5ca5926258e4f0b5041fb2c15b1c2f904c4adb
539 Author: Sage Weil <sage@inktank.com>
540 Date: Mon Jan 21 13:57:59 2013 -0800
542 osd: set PULL subop cost to size of requested data
544 Signed-off-by: Sage Weil <sage@inktank.com>
545 (cherry picked from commit a1bf8220e545f29b83d965f07b1abfbea06238b3)
547 commit 10651e4f500d7b55d8c689a10a61d2239b3ecd26
548 Author: Sage Weil <sage@inktank.com>
549 Date: Mon Jan 21 13:57:38 2013 -0800
551 osd: use Message::get_cost() function for queueing
553 The data payload is a decent proxy for cost in most cases, but not all.
555 Signed-off-by: Sage Weil <sage@inktank.com>
556 (cherry picked from commit e8e0da1a577e24cd4aad71fb94d8b244e2ac7300)
558 commit 9735c6b163f4d226d8de6508d5c1534d18f1c300
559 Author: Sage Weil <sage@inktank.com>
560 Date: Mon Jan 21 13:25:21 2013 -0800
562 osd: debug msg prio, cost, latency
564 Signed-off-by: Sage Weil <sage@inktank.com>
565 (cherry picked from commit bec96a234c160bebd9fd295df5b431dc70a2cfb3)
567 commit c48279da7ad98013ce97eab89c17fe9fae1ba866
568 Author: Sage Weil <sage@inktank.com>
569 Date: Mon Jan 21 21:05:00 2013 -0800
571 filestore: filestore_queue_max_ops 500 -> 50
573 Having a deep queue limits the effectiveness of the priority queues
574 above by adding additional latency.
576 Signed-off-by: Sage Weil <sage@inktank.com>
577 (cherry picked from commit 40654d6d53436c210b2f80911217b044f4d7643a)
579 commit f47b2e8b607cc0d56a42ec7b1465ce6b8c0ca68c
580 Author: Sage Weil <sage@inktank.com>
581 Date: Mon Jan 21 20:00:26 2013 -0800
583 osd: target transaction size 300 -> 30
585 Small transactions make pg removal nicer to the op queue. It also slows
586 down PG deletion a bit, which may exacerbate the PG resurrection case
587 until #3884 is addressed.
589 At least on user reported this fixed an osd that kept failing due to
590 an internal heartbeat failure.
592 Signed-off-by: Sage Weil <sage@inktank.com>
593 (cherry picked from commit 1233e8617098766c95100aa9a6a07db1a688e290)
595 commit 4947f0efadf9ef209d02fd17f5f86b9a7d6523ef
596 Author: Sage Weil <sage@inktank.com>
597 Date: Mon Jan 21 19:55:26 2013 -0800
599 os/FileStore: allow filestore_queue_max_{ops,bytes} to be adjusted at runtime
601 The 'committing' ones too.
603 Signed-off-by: Sage Weil <sage@inktank.com>
604 (cherry picked from commit cfe4b8519363f92f84f724a812aa41257402865f)
606 commit ad6e6c91f61c092bfc9f88b788ccbee6438fd40b
607 Author: Sage Weil <sage@inktank.com>
608 Date: Sat Jan 19 22:06:27 2013 -0800
610 osd: make osd_max_backfills dynamically adjustable
612 Signed-off-by: Sage Weil <sage@inktank.com>
613 (cherry picked from commit 101955a6b8bfdf91f4229f4ecb5d5b3da096e160)
615 commit 939b1855245bc9cb31f5762027f2ed3f2317eb55
616 Author: Sage Weil <sage@inktank.com>
617 Date: Sat Jan 19 18:28:35 2013 -0800
619 osd: make OSD a config observer
621 Signed-off-by: Sage Weil <sage@inktank.com>
622 (cherry picked from commit 9230c863b3dc2bdda12c23202682a84c48f070a1)
628 commit b0f27a8f81feb401407bed784bf5d4d799998ee0
629 Author: Dan Mick <dan.mick@inktank.com>
630 Date: Tue Jan 8 11:21:22 2013 -0800
632 librbd: Allow get_lock_info to fail
634 If the lock class isn't present, EOPNOTSUPP is returned for lock calls
635 on newer OSDs, but sadly EIO on older; we need to treat both as
636 acceptable failures for RBD images. rados lock list will still fail.
640 Signed-off-by: Dan Mick <dan.mick@inktank.com>
641 Reviewed-by: Sage Weil <sage@inktank.com>
642 (cherry picked from commit 4483285c9fb16f09986e2e48b855cd3db869e33c)
644 commit 022a5254b4fac3f76220abdde2a2e81de33cb8dc
645 Author: Sage Weil <sage@inktank.com>
646 Date: Fri Jan 4 13:00:56 2013 -0800
648 osd: drop newlines from event descriptions
650 These produce extra newlines in the log.
652 Signed-off-by: Sage Weil <sage@inktank.com>
653 Reviewed-by: Samuel Just <sam.just@inktank.com>
654 (cherry picked from commit 9a1f574283804faa6dbba9165a40558e1a6a1f13)
656 commit ebc93a878c8b0697004a619d6aa957a80b8b7e35
657 Author: Samuel Just <sam.just@inktank.com>
658 Date: Fri Jan 18 14:35:51 2013 -0800
660 OSD: do deep_scrub for repair
662 Signed-off-by: Samuel Just <sam.just@inktank.com>
663 Reviewed-by: David Zafman <david.zafman@inktank.com>
664 (cherry picked from commit 0cb760f31b0cb26f022fe8b9341e41cd5351afac)
666 commit 32527fa3eb48a7d7d5d67c39bfa05087dbc0e41b
667 Author: Samuel Just <sam.just@inktank.com>
668 Date: Mon Jan 14 12:52:04 2013 -0800
670 ReplicatedPG: ignore snap link info in scrub if nlinks==0
672 links==0 implies that the replica did not sent snap link information.
674 Signed-off-by: Samuel Just <sam.just@inktank.com>
675 (cherry picked from commit 70c3512037596a42ba6eb5eb7f96238843095db9)
677 commit 13e42265db150b19511a5a618c7a95ad801290c8
678 Author: Sage Weil <sage@inktank.com>
679 Date: Fri Jan 11 12:25:22 2013 -0800
681 osd/PG: fix osd id in error message on snap collection errors
683 Signed-off-by: Sage Weil <sage@inktank.com>
684 (cherry picked from commit 381e25870f26fad144ecc2fb99710498e3a7a1d4)
686 commit e3b6191fc45c7d2c27ec75c867be822a6da17e9a
687 Author: Sage Weil <sage@inktank.com>
688 Date: Wed Jan 9 22:34:12 2013 -0800
690 osd/ReplicatedPG: validate ino when scrubbing snap collections
692 Signed-off-by: Sage Weil <sage@inktank.com>
693 (cherry picked from commit 665577a88b98390b9db0f9991836d10ebdd8f4cf)
695 commit 353b7341caff86f936a429669de52e6949a89c2b
696 Author: Samuel Just <sam.just@inktank.com>
697 Date: Wed Jan 9 16:41:40 2013 -0800
699 ReplicatedPG: compare nlinks to snapcolls
701 nlinks gives us the number of hardlinks to the object.
702 nlinks should be 1 + snapcolls.size(). This will allow
703 us to detect links which remain in an erroneous snap
706 Signed-off-by: Samuel Just <sam.just@inktank.com>
707 (cherry picked from commit e65ea70ea64025fbb0709ee8596bb2878be0bbdc)
709 commit 33d5cfc8c080a270d65275f8e010a6468c77381a
710 Author: Samuel Just <sam.just@inktank.com>
711 Date: Thu Jan 10 15:35:10 2013 -0800
713 ReplicatedPG/PG: check snap collections during _scan_list
715 During _scan_list check the snapcollections corresponding to the
716 object_info attr on the object. Report inconsistencies during
719 Signed-off-by: Samuel Just <sam.just@inktank.com>
720 (cherry picked from commit 57352351bb86e0ae9f64f9ba0d460c532d882de6)
722 commit bea783bd722d862a5018477a637c843fe4b18a58
723 Author: Samuel Just <sam.just@inktank.com>
724 Date: Wed Jan 9 11:53:52 2013 -0800
726 osd_types: add nlink and snapcolls fields to ScrubMap::object
728 Signed-off-by: Samuel Just <sam.just@inktank.com>
729 (cherry picked from commit b85687475fa2ec74e5429d92ee64eda2051a256c)
731 commit 0c48407bf46b39b2264a7be14e9d3caa2c1e5875
732 Author: Samuel Just <sam.just@inktank.com>
733 Date: Thu Jan 3 20:16:50 2013 -0800
735 PG: move auth replica selection to helper in scrub
737 Signed-off-by: Samuel Just <sam.just@inktank.com>
738 (cherry picked from commit 39bc65492af1bf1da481a8ea0a70fe7d0b4b17a3)
740 commit c3433ce60ec3683217d8b4cd2b6e75fb749af2c6
741 Author: Sage Weil <sage@inktank.com>
742 Date: Mon Jan 14 18:23:52 2013 -0800
744 mon: note scrub errors in health summary
746 Signed-off-by: Sage Weil <sage@inktank.com>
747 (cherry picked from commit 8e33a8b9e1fef757bbd901d55893e9b84ce6f3fc)
749 commit 90c6edd0155b327c48a5b178d848d9e5839bd928
750 Author: Sage Weil <sage@inktank.com>
751 Date: Mon Jan 14 18:31:06 2013 -0800
753 osd: fix rescrub after repair
755 We were rescrubbing if INCONSISTENT is set, but that is now persistent.
756 Add a new scrub_after_recovery flag that is reset on each peering interval
757 and set that when repair encounters errors.
759 Signed-off-by: Sage Weil <sage@inktank.com>
760 (cherry picked from commit a586966a3cfb10b5ffec0e9140053a7e4ff105d2)
762 commit 0696cf57283e6e9a3500c56ca5fc9f981475ca26
763 Author: Sage Weil <sage@inktank.com>
764 Date: Mon Jan 14 18:22:02 2013 -0800
766 osd: note must_scrub* flags in PG operator<<
768 Signed-off-by: Sage Weil <sage@inktank.com>
769 (cherry picked from commit d56af797f996ac92bf4e0886d416fd358a2aa08e)
771 commit 1541ffe4bec6cce607c505271ff074fd0a292d30
772 Author: Sage Weil <sage@inktank.com>
773 Date: Mon Jan 14 18:21:46 2013 -0800
775 osd: based INCONSISTENT pg state on persistent scrub errors
777 This makes the state persistent across PG peering and OSD restarts.
779 This has the side-effect that, on recovery, we rescrub any PGs marked
780 inconsistent. This is new behavior!
782 Signed-off-by: Sage Weil <sage@inktank.com>
783 (cherry picked from commit 2baf1253eed630a7c4ae4cb43aab6475efd82425)
785 commit 609101255c81d977072b2ab741ac47167d9b1b16
786 Author: Sage Weil <sage@inktank.com>
787 Date: Mon Jan 14 18:20:29 2013 -0800
789 osd: fix scrub scheduling for 0.0
791 The initial value for pair<utime_t,pg_t> can match pg 0.0, preventing it
792 from being manually scrubbed. Fix!
794 Signed-off-by: Sage Weil <sage@inktank.com>
795 (cherry picked from commit 26a63df97b2a12fd1a7c1e3cc9ccd34ca2ef9834)
797 commit 0961a3a85c286a31ec2e8bba23217bbd3974572c
798 Author: Sage Weil <sage@inktank.com>
799 Date: Sun Jan 13 23:03:01 2013 -0800
801 osd: note last_clean_scrub_stamp, last_scrub_errors
803 Signed-off-by: Sage Weil <sage@inktank.com>
804 (cherry picked from commit 389bed5d338cf32ab14c9fc2abbc7bcc386b8a28)
806 commit 8d823045538bf4c51506e349b5c6705fd76450f8
807 Author: Sage Weil <sage@inktank.com>
808 Date: Sun Jan 13 22:59:39 2013 -0800
810 osd: add num_scrub_errors to object_stat_t
812 Signed-off-by: Sage Weil <sage@inktank.com>
813 (cherry picked from commit 2475066c3247774a2ad048a2e32968e47da1b0f5)
815 commit 3a1cd6e07b4e2a4714de159f69afd689495e2927
816 Author: Sage Weil <sage@inktank.com>
817 Date: Sun Jan 13 22:43:35 2013 -0800
819 osd: add last_clean_scrub_stamp to pg_stat_t, pg_history_t
821 Signed-off-by: Sage Weil <sage@inktank.com>
822 (cherry picked from commit d738328488de831bf090f23e3fa6d25f6fa819df)
824 commit 7e5a899bdcf6c08a5f6f5c98cd2fff7fa2dacaca
825 Author: Sage Weil <sage@inktank.com>
826 Date: Sun Jan 13 22:56:14 2013 -0800
828 osd: fix object_stat_sum_t dump signedness
830 Signed-off-by: Sage Weil <sage@inktank.com>
831 (cherry picked from commit 6f6a41937f1bd05260a8d70b4c4a58ecadb34a2f)
833 commit e252a313d465006d3fe4db97939ad307ebe91c71
834 Author: Sage Weil <sage@inktank.com>
835 Date: Sun Jan 13 22:04:58 2013 -0800
837 osd: change scrub min/max thresholds
839 The previous 'osd scrub min interval' was mostly meaningless and useless.
840 Meanwhile, the 'osd scrub max interval' would only trigger a scrub if the
841 load was sufficiently low; if it was high, the PG might *never* scrub.
843 Instead, make the 'min' what the max used to be. If it has been more than
844 this many seconds, and the load is low, scrub. And add an additional
845 condition that if it has been more than the max threshold, scrub the PG
846 no matter what--regardless of the load.
848 Note that this does not change the default scrub interval for less-loaded
849 clusters, but it *does* change the meaning of existing config options.
852 Signed-off-by: Sage Weil <sage@inktank.com>
853 (cherry picked from commit 299548024acbf8123a4e488424c06e16365fba5a)
859 commit 33aa64eee34f4759f6000130de4d1306de49d087
860 Author: Sage Weil <sage@inktank.com>
861 Date: Sun Jan 13 20:27:59 2013 -0800
863 osd/PG: remove useless osd_scrub_min_interval check
865 This was already a no-op: we don't call PG::scrub_sched() unless it has
866 been osd_scrub_max_interval seconds since we last scrubbed. Unless we
867 explicitly requested in, in which case we don't want this check anyway.
869 Signed-off-by: Sage Weil <sage@inktank.com>
870 (cherry picked from commit 16d67c798b6f752a6e03084bafe861396b86baae)
872 commit fdd0c1ec3519376980a205b94e65187833634e2e
873 Author: Sage Weil <sage@inktank.com>
874 Date: Sun Jan 13 20:25:39 2013 -0800
876 osd: move scrub schedule random backoff to seperate helper
878 Separate this from the load check, which will soon vary dependon on the
881 Signed-off-by: Sage Weil <sage@inktank.com>
882 (cherry picked from commit a148120776d0930b265411332a60e93abfbf0423)
884 commit 9ffbe268f785e1a74c0d893735117edb7a3ef377
885 Author: Sage Weil <sage@inktank.com>
886 Date: Sat Jan 12 09:18:38 2013 -0800
888 osd/PG: trigger scrub via scrub schedule, must_ flags
890 When a scrub is requested, flag it and move it to the front of the
891 scrub schedule instead of immediately queuing it. This avoids
892 bypassing the scrub reservation framework, which can lead to a heavier
893 impact on performance.
895 Signed-off-by: Sage Weil <sage@inktank.com>
896 (cherry picked from commit 62ee6e099a8e4873287b54f9bba303ea9523d040)
898 commit cffb1b22d5df7300ec411d2b620bf3c4a08351cd
899 Author: Sage Weil <sage@inktank.com>
900 Date: Sat Jan 12 09:15:16 2013 -0800
902 osd/PG: introduce flags to indicate explicitly requested scrubs
904 Signed-off-by: Sage Weil <sage@inktank.com>
905 (cherry picked from commit 1441095d6babfacd781929e8a54ed2f8a4444467)
907 commit 438e3dfc88bfdc8eb36b5b5f7b728b2610476724
908 Author: Sage Weil <sage@inktank.com>
909 Date: Sat Jan 12 09:14:01 2013 -0800
911 osd/PG: move scrub schedule registration into a helper
913 Simplifies callers, and will let us easily modify the decision of when
914 to schedule the PG for scrub.
916 Signed-off-by: Sage Weil <sage@inktank.com>
917 (cherry picked from commit 796907e2159371f84a16cbd35f6caa8ac868acf6)
919 commit acb47e4d7dc9682937984661a9d754131d806630
920 Author: Sage Weil <sage@inktank.com>
921 Date: Fri Jan 18 12:14:48 2013 -0800
923 os/FileStore: only flush inline if write is sufficiently large
925 Honor filestore_flush_min in the inline flush case.
928 Signed-off-by: Sage Weil <sage@inktank.com>
929 Reviewed-by: Samuel Just <sam.just@inktank.com>
930 (cherry picked from commit 49726dcf973c38c7313ab78743b45ccc879671ea)
932 commit 15a1ced859629c361da127799b05620bee84c9a8
933 Author: Sage Weil <sage@inktank.com>
934 Date: Fri Jan 18 12:14:40 2013 -0800
936 os/FileStore: fix compile when sync_file_range is missing;
938 If sync_file_range is not present, we always close inline, and flush
941 Fixes compile on ancient platforms like RHEL5.8.
944 Signed-off-by: Sage Weil <sage@inktank.com>
945 Reviewed-by: Samuel Just <sam.just@inktank.com>
946 (cherry picked from commit 8ddb55d34c72e6df1023cf427cbd41f3f98da402)
948 commit 9dddb9d855e6d5fd804b54bff1f726c1d2fb566c
949 Author: Sage Weil <sage@inktank.com>
950 Date: Fri Jan 18 15:23:22 2013 -0800
952 osd: set pg removal transactions based on configurable
954 Use the osd_target_transaction_size knob, and gracefully tolerate bogus
957 Signed-off-by: Sage Weil <sage@inktank.com>
958 (cherry picked from commit 5e00af406b89c9817e9a429f92a05ca9c29b19c3)
960 commit c30d231e40a17c3fb08d1db5e01133466170e90c
961 Author: Sage Weil <sage@inktank.com>
962 Date: Fri Jan 18 15:30:06 2013 -0800
964 osd: make pg removal thread more friendly
966 For a large PG these are saturating the filestore and journal queues. Do
967 them synchronously to make them more friendly. They don't need to be fast.
969 Signed-off-by: Sage Weil <sage@inktank.com>
970 (cherry picked from commit 4712e984d3f62cdf51ea67da8197eed18a5983dd)
972 commit b2bc4b95fefaeb0cfc31ce0bc95b77062d0777c7
973 Author: Sage Weil <sage@inktank.com>
974 Date: Fri Jan 18 15:27:24 2013 -0800
976 os: move apply_transactions() sync wrapper into ObjectStore
978 This has nothing to do with the backend implementation.
980 Signed-off-by: Sage Weil <sage@inktank.com>
981 (cherry picked from commit bc994045ad67fb70c7a0457b8cd29273dd5d1654)
983 commit 6d161b57979246ddea4e6309e0e489ab729eec4b
984 Author: Sage Weil <sage@inktank.com>
985 Date: Fri Jan 18 15:28:24 2013 -0800
987 os: add apply_transaction() variant that takes a sequencer
989 Also, move the convenience wrappers into the interface and funnel through
990 a single implementation.
992 Signed-off-by: Sage Weil <sage@inktank.com>
993 (cherry picked from commit f6c69c3f1ac35546b90315fff625993ba5cd8c07)
995 commit c5fe0965572c074a2a33660719ce3222d18c1464
996 Author: Sage Weil <sage@inktank.com>
997 Date: Sun Jan 20 16:11:10 2013 -0800
999 osd: calculate initial PG mapping from PG's osdmap
1001 The initial values of up/acting need to be based on the PG's osdmap, not
1002 the OSD's latest. This can cause various confusion in
1003 pg_interval_t::check_new_interval() when calling OSDMap methods due to the
1004 up/acting OSDs not existing yet (for example).
1007 Reported-by: Jens Kristian S?gaard <jens@mermaidconsulting.dk>
1008 Tested-by: Jens Kristian S?gaard <jens@mermaidconsulting.dk>
1009 Signed-off-by: Sage Weil <sage@inktank.com>
1010 Reviewed-by: Samuel Just <sam.just@inktank.com>
1011 (cherry picked from commit 17160843d0c523359d8fa934418ff2c1f7bffb25)
1013 commit 6008b1d8e4587d5a3aea60684b1d871401496942
1014 Author: Sage Weil <sage@inktank.com>
1015 Date: Thu Jan 17 15:01:35 2013 -0800
1017 osdmap: make replica separate in default crush map configurable
1019 Add 'osd crush chooseleaf type' option to control what the default
1020 CRUSH rule separates replicas across. Default to 1 (host), and set it
1024 Signed-off-by: Sage Weil <sage@inktank.com>
1025 Reviewed-by: Greg Farnum <greg@inktank.com>
1026 (cherry picked from commit c236a51a8040508ee893e4c64b206e40f9459a62)
1028 commit 5fb77bf1d1b241b4f9c1fe9e57288bbc84d8d97d
1029 Author: Sage Weil <sage@inktank.com>
1030 Date: Wed Jan 16 14:09:53 2013 -0800
1032 ceph: adjust crush tunables via 'ceph osd crush tunables <profile>'
1034 Make it easy to adjust crush tunables. Create profiles:
1036 legacy: the legacy values
1037 argonaut: the argonaut defaults, and what is supported.. legacy! (*(
1038 bobtail: best that bobtail supports
1039 optimal: the current optimal values
1040 default: the current default values
1042 * In actuality, argonaut supports some of the tunables, but it doesn't
1043 say so via the feature bits.
1045 Signed-off-by: Sage Weil <sage@inktank.com>
1046 Reviewed-by: Samuel Just <sam.just@inktank.com>
1047 Reviewed-by: Dan Mick <dan.mick@inktank.com>
1048 (cherry picked from commit 19ee23111585f15a39ee2907fa79e2db2bf523f0)
1050 commit 8c0d702e6f2ba0ed0fe31c06c7a028260ae08e42
1051 Author: Sage Weil <sage@inktank.com>
1052 Date: Fri Dec 28 17:20:43 2012 -0800
1054 msg/Pipe: use state_closed atomic_t for _lookup_pipe
1056 We shouldn't look at Pipe::state in SimpleMessenger::_lookup_pipe() without
1057 holding pipe_lock. Instead, use an atomic that we set to non-zero only
1058 when transitioning to the terminal STATE_CLOSED state.
1060 Signed-off-by: Sage Weil <sage@inktank.com>
1061 (cherry picked from commit 82f8bcddb5fa09913eb477ee26c71d6b4bb8d97c)
1063 commit 8e0359c3e586c0edcce769c8ed1a03444a521165
1064 Author: Sage Weil <sage@inktank.com>
1065 Date: Sun Dec 23 13:43:15 2012 -0800
1067 msgr: inject delays at inconvenient times
1069 Exercise some rare races by injecting delays before taking locks
1070 via the 'ms inject internal delays' option.
1072 Signed-off-by: Sage Weil <sage@inktank.com>
1073 (cherry picked from commit a5d692a7b9b4bec2c27993ca37aa3fec4065292b)
1075 commit 34e2d4024700f633c2c586265efb61537342db18
1076 Author: Sage Weil <sage@inktank.com>
1077 Date: Sun Dec 23 09:22:18 2012 -0800
1079 msgr: fix race on Pipe removal from hash
1081 When a pipe is faulting and shutting down, we have to drop pipe_lock to
1082 take msgr lock and then remove the entry. The Pipe in this case will
1083 have STATE_CLOSED. Handle this case in all places we do a lookup on
1084 the rank_pipe hash so that we effectively ignore entries that are
1087 This fixes a race introduced by the previous commit where we won't use
1088 the CLOSED pipe and try to register a new one, but the old one is still
1093 Signed-off-by: Sage Weil <sage@inktank.com>
1094 (cherry picked from commit e99b4a307b4427945a4eb5ec50e65d6239af4337)
1096 commit ae1882e7efc91b770ac0ac8682ee6c5792a63a93
1097 Author: Sage Weil <sage@inktank.com>
1098 Date: Sun Dec 23 09:19:05 2012 -0800
1100 msgr: don't queue message on closed pipe
1102 If we have a con that refs a pipe but it is closed, don't use it. If
1103 the ref is still there, it is only because we are racing with fault()
1104 and it is about to (or just was) be detached. Either way,
1106 Signed-off-by: Sage Weil <sage@inktank.com>
1107 (cherry picked from commit 6339c5d43974f4b495f15d199e01a141e74235f5)
1109 commit 373f1671b6cb64dba5a9172967b27177515be1fd
1110 Author: Sage Weil <sage@inktank.com>
1111 Date: Sat Dec 22 21:24:52 2012 -0800
1113 msgr: atomically queue first message with connect_rank
1115 Atomically queue the first message on the new pipe, without dropping
1116 and retaking pipe_lock.
1118 Signed-off-by: Sage Weil <sage@inktank.com>
1119 (cherry picked from commit 7bf0b0854d1f2706a3a2302bcbf92dd5c8c888ef)
1121 commit 82f22b38c5dc0b636574679ba1fee1b36a3c0478
1122 Author: Samuel Just <sam.just@inktank.com>
1123 Date: Thu Jan 10 11:06:02 2013 -0800
1125 config_opts.h: default osd_recovery_delay_start to 0
1127 This setting was intended to prevent recovery from overwhelming peering traffic
1128 by delaying the recovery_wq until osd_recovery_delay_start seconds after pgs
1129 stop being added to it. This should be less necessary now that recovery
1130 messages are sent with strictly lower priority then peering messages.
1132 Signed-off-by: Samuel Just <sam.just@inktank.com>
1133 Reviewed-by: Gregory Farnum <greg@inktank.com>
1134 (cherry picked from commit 44625d4460f61effe2d63d8280752f10f159e7b4)
1136 commit 81e8bb55e28384048fd82116a791a65ca52ef999
1137 Author: Sage Weil <sage@inktank.com>
1138 Date: Wed Jan 16 21:19:18 2013 -0800
1140 osdmaptool: more fix cli test
1142 Signed-off-by: Sage Weil <sage@inktank.com>
1143 (cherry picked from commit b0162fab3d927544885f2b9609b9ab3dc4aaff74)
1145 commit 2b5b2657579abdf5b1228f4c5c5ac8cec3706726
1146 Author: Sage Weil <sage@inktank.com>
1147 Date: Wed Jan 16 21:10:26 2013 -0800
1149 osdmaptool: fix cli test
1151 Signed-off-by: Sage Weil <sage@inktank.com>
1152 (cherry picked from commit 5bd8765c918174aea606069124e43c480c809943)
1154 commit f739d1238a8a67598c037b6e2ed5d539a2d79996
1155 Author: Samuel Just <sam.just@inktank.com>
1156 Date: Wed Jan 16 14:21:47 2013 -0800
1158 osdmaptool: allow user to specify pool for test-map-object
1162 Signed-off-by: Samuel Just <sam.just@inktank.com>
1163 Reviewed-by: Gregory Farnum <greg@inktank.com>
1164 (cherry picked from commit 85eb8e382a26dfc53df36ae1a473185608b282aa)
1166 commit 00759ee08f5dc62cbe4f237399f298472f6d8f4a
1167 Author: David Zafman <david.zafman@inktank.com>
1168 Date: Wed Jan 16 12:41:16 2013 -0800
1170 rados.cc: fix rmomapkey usage: val not needed
1172 Signed-off-by: David Zafman <david.zafman@inktank.com>
1173 Reviewed-by: Samuel Just <samuel.just@inktank.com>
1174 (cherry picked from commit 625c3cb9b536a0cff7249b8181b7a4f09b1b4f4f)
1176 commit 06b3270f679be496df41810dacf863128b0cfcaa
1177 Author: Samuel Just <sam.just@inktank.com>
1178 Date: Tue Jan 15 21:27:23 2013 -0800
1180 librados.hpp: fix omap_get_vals and omap_get_keys comments
1182 We list keys greater than start_after.
1184 Signed-off-by: Samuel Just <sam.just@inktank.com>
1185 Reviewed-by: David Zafman <david.zafman@inktank.com>
1186 (cherry picked from commit 3f0ad497b3c4a5e9bef61ecbae5558ae72d4ce8b)
1188 commit 75072965201380aa55a8e15f9db4ccaf4d34d954
1189 Author: Samuel Just <sam.just@inktank.com>
1190 Date: Tue Jan 15 21:26:22 2013 -0800
1192 rados.cc: use omap_get_vals_by_keys in getomapval
1195 Signed-off-by: Samuel Just <sam.just@inktank.com>
1196 Reviewed-by: David Zafman <david.zafman@inktank.com>
1197 (cherry picked from commit cb5e2be418924cf8b2c6a6d265a7a0327f08d00a)
1199 commit a3c2980fccfe95b7d094a7c93945437c3911b858
1200 Author: Samuel Just <sam.just@inktank.com>
1201 Date: Tue Jan 15 21:24:50 2013 -0800
1203 rados.cc: fix listomapvals usage: key,val are not needed
1206 Signed-off-by: Samuel Just <sam.just@inktank.com>
1207 Reviewed-by: David Zafman <david.zafman@inktank.com>
1208 (cherry picked from commit 44c45e520cc2e60c6c803bb245edb9330bff37e4)
1210 commit 20b27a1ce71c379a3b2a29d282dc0689a3a0df46
1211 Author: Yehuda Sadeh <yehuda@inktank.com>
1212 Date: Wed Jan 16 15:01:47 2013 -0800
1214 rgw: copy object should not copy source acls
1217 Backport: argonaut, bobtail
1219 When using the S3 api and x-amz-metadata-directive is
1220 set to COPY we used to copy complete metadata of source
1221 object. However, this shouldn't include the source ACLs.
1223 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
1224 (cherry picked from commit 37dbf7d9df93dd0e92019be31eaa1a19dd9569c7)
1226 commit 3293b31b44c9adad2b5e37da9d5342a6e4b72ade
1227 Author: Samuel Just <sam.just@inktank.com>
1228 Date: Fri Jan 11 11:02:15 2013 -0800
1230 OSD: only trim up to the oldest map still in use by a pg
1232 map_cache.cached_lb() provides us with a lower bound across
1233 all pgs for in-use osdmaps. We cannot trim past this since
1234 those maps are still in use.
1238 Signed-off-by: Samuel Just <sam.just@inktank.com>
1239 Reviewed-by: Sage Weil <sage@inktank.com>
1240 Reviewed-by: Greg Farnum <greg@inktank.com>
1241 (cherry picked from commit 66eb93b83648b4561b77ee6aab5b484e6dba4771)
1243 commit 898a4b19ecc6fffc33feb198f37182ec0a6e77e9
1244 Author: Sage Weil <sage@inktank.com>
1245 Date: Mon Jan 14 08:15:02 2013 -0800
1247 Revert "osdmap: spread replicas across hosts with default crush map"
1249 This reverts commit 503917f0049d297218b1247dc0793980c39195b3.
1251 This breaks vstart and teuthology configs. A better fix is coming.
1253 commit 55b7dd3248f35929ea097525798e8667fafbf161
1254 Author: Joao Eduardo Luis <joao.luis@inktank.com>
1255 Date: Thu Jan 10 18:54:12 2013 +0000
1257 mon: OSDMonitor: don't output to stdout in plain text if json is specified
1261 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
1262 Reviewed-by: Sage Weil <sage@inktank.com>
1263 (cherry picked from commit 410906e04936c935903526f26fb7db16c412a711)
1265 commit 015a454a0c046cb678991cc4f4d53fb58c41dbe4
1266 Author: Sage Weil <sage@inktank.com>
1267 Date: Fri Jan 11 17:23:22 2013 -0800
1269 osdmap: spread replicas across hosts with default crush map
1271 This is more often the case than not, and we don't have a good way to
1272 magically know what size of cluster the user will be creating. Better to
1273 err on the side of doing the right thing for more people.
1276 Signed-off-by: Sage Weil <sage@inktank.com>
1277 Reviewed-by: Greg Farnum <greg@inktank.com>
1278 (cherry picked from commit 7ea5d84fa3d0ed3db61eea7eb9fa8dbee53244b6)
1280 commit d882d053927c319274be38a247f2beabb4e06b64
1281 Author: Samuel Just <sam.just@inktank.com>
1282 Date: Wed Jan 9 19:17:23 2013 -0800
1284 ReplicatedPG: fix snapdir trimming
1286 The previous logic was both complicated and not correct. Consequently,
1287 we have been tending to drop snapcollection links in some cases. This
1288 has resulted in clones incorrectly not being trimmed. This patch
1289 replaces the logic with something less efficient but hopefully a bit
1292 Signed-off-by: Samuel Just <sam.just@inktank.com>
1293 Reviewed-by: Sage Weil <sage@inktank.com>
1294 (cherry picked from commit 0f42c37359d976d1fe90f2d3b877b9b0268adc0b)