1 commit 92b1e398576d55df8e5888dd1a9545ed3fd99532
2 Author: Gary Lowell <gary.lowell@inktank.com>
3 Date: Wed Jun 5 11:10:05 2013 -0700
7 commit 7d549cb82ab8ebcf1cc104fc557d601b486c7635
8 Author: Sage Weil <sage@inktank.com>
9 Date: Tue Jun 4 10:42:13 2013 -0700
11 os/LevelDBStore: only remove logger if non-null
13 Signed-off-by: Sage Weil <sage@inktank.com>
14 (cherry picked from commit ce67c58db7d3e259ef5a8222ef2ebb1febbf7362)
17 commit 8544ea751884617616addc17b4467b9a86bd9d8a
18 Author: Josh Durgin <josh.durgin@inktank.com>
19 Date: Mon Jun 3 15:57:23 2013 -0700
21 test_librbd: use correct type for varargs snap test
23 uint64_t is passed in, but int was extracted. This fails on 32-bit builds.
26 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
27 (cherry picked from commit 17029b270dee386e12e5f42c2494a5feffd49b08)
29 commit b226e117b5a72c3b04b74aec50a9198601f3730b
30 Author: Sage Weil <sage@inktank.com>
31 Date: Sun Jun 2 18:07:34 2013 -0700
33 os/LevelDBStore: fix merge loop
35 We were double-incrementing p, both in the for statement and in the
36 body. While we are here, drop the unnecessary else's.
38 Signed-off-by: Sage Weil <sage@inktank.com>
39 (cherry picked from commit eb6d5fcf994d2a25304827d7384eee58f40939af)
41 commit d017afb5b075a1958cb19213dd95e41d54065e45
42 Author: Sage Weil <sage@inktank.com>
43 Date: Sun Jun 2 17:27:10 2013 -0700
45 msgr: add get_messenger() to Connection
47 This was part of commit 27381c0c6259ac89f5f9c592b4bfb585937a1cfc.
49 Signed-off-by: Sage Weil <sage@inktank.com>
51 commit ffb87918fa7b829a5199eec08804dc540a819bf2
52 Author: Sage Weil <sage@inktank.com>
53 Date: Fri May 31 17:09:19 2013 -0700
55 mon: start lease timer from peon_init()
59 - leader wins, peons lose
60 - leader sees it is too far behind on paxos and bootstraps
61 - leader tries to sync with someone, waits for a quorum of the others
62 - peons sit around forever waiting
64 The problem is that they never time out because paxos never issues a lease,
65 which is the normal timeout that lets them detect a leader failure.
67 Avoid this by starting the lease timeout as soon as we lose the election.
68 The timeout callback just does a bootstrap and does not rely on any other
71 I see one possible danger here: there may be some "normal" cases where the
72 leader takes a long time to issue its first lease that we currently
73 tolerate, but won't with this new check in place. I hope that raising
74 the lease interval/timeout or reducing the allowed paxos drift will make
75 that a non-issue. If it is problematic, we will need a separate explicit
76 "i am alive" from the leader while it is getting ready to issue the lease
77 to prevent a live-lock.
79 Backport: cuttlefish, bobtail
80 Signed-off-by: Sage Weil <sage@inktank.com>
81 Reviewed-by: Greg Farnum <greg@inktank.com>
82 (cherry picked from commit f1ccb2d808453ad7ef619c2faa41a8f6e0077bd9)
84 commit 38f8d850d35500e3d8751cd14c5cdaaff682c7d7
85 Author: Sage Weil <sage@inktank.com>
86 Date: Thu May 30 22:52:21 2013 -0700
88 mon: discard messages from disconnected clients
90 If the client is not connected, discard the message. They will
91 reconnect and resend anyway, so there is no point in processing it
92 twice (now and later).
94 Signed-off-by: Sage Weil <sage@inktank.com>
95 Reviewed-by: Greg Farnum <greg@inktank.com>
96 (cherry picked from commit fb3cd0c2a8f27a1c8d601a478fd896cc0b609011)
98 commit 36d948981685114d2fe807f480c19aade7497194
99 Author: Sage Weil <sage@inktank.com>
100 Date: Wed May 22 08:13:21 2013 -0700
102 msgr: add Messenger reference to Connection
104 This allows us to get the messenger associated with a connection.
106 Signed-off-by: Sage Weil <sage@inktank.com>
107 (cherry picked from commit 92a558bf0e5fee6d5250e1085427bff22fe4bbe4)
109 commit 61135964419ecf5165366724d064b623b517fb4e
110 Author: Sage Weil <sage@inktank.com>
111 Date: Thu May 30 15:59:49 2013 -0700
113 mon/Paxos: adjust trimming defaults up; rename options
115 - trim more at a time (by an order of magnitude)
116 - rename fields to paxos_trim_{min,max}; only trim when there are min items
117 that are trimmable, and trim at most max items at a time.
118 - adjust the paxos_service_trim_{min,max} values up by a factor of 2.
120 Since we are compacting every time we trim, adjusting these up mean less
121 frequent compactions and less overall work for the monitor.
123 Signed-off-by: Sage Weil <sage@inktank.com>
124 Reviewed-by: Greg Farnum <greg@inktank.com>
125 (cherry picked from commit 6b8e74f0646a7e0d31db24eb29f3663fafed4ecc)
127 commit 2dc402815f71204cfe592cfb3d6758486d84166d
128 Author: Sage Weil <sage@inktank.com>
129 Date: Wed May 8 16:42:24 2013 -0700
131 common/Preforker: fix warnings
133 Signed-off-by: Sage Weil <sage@inktank.com>
134 (cherry picked from commit a284c9ece85f11d020d492120be66a9f4c997416)
136 commit 48ee9283de288c101d3387cc48df6eb8ea889fb7
137 Author: Sage Weil <sage@inktank.com>
138 Date: Thu May 30 15:53:35 2013 -0700
140 fix test users of LevelDBStore
144 Signed-off-by: Sage Weil <sage@inktank.com>
145 (cherry picked from commit 446e0770c77de5d72858dcf7a95c5b19f642cf98)
147 commit 3372235d307172b404a48e4c4d24702a96116d00
148 Author: Sage Weil <sage@inktank.com>
149 Date: Thu May 30 21:43:50 2013 -0700
151 mon: destroy MonitorDBStore before g_ceph_context
153 Put it on the heap so that we can destroy it before the g_ceph_context
154 cct that it references. This fixes a crash like
156 *** Caught signal (Segmentation fault) **
158 ceph version 0.63-204-gcf9aa7a (cf9aa7a0037e56eada8b3c1bb59d59d0bfe7bba5)
159 1: ceph-mon() [0x59932a]
160 2: (()+0xfcb0) [0x4e41cb0]
161 3: (Mutex::Lock(bool)+0x1b) [0x6235bb]
162 4: (PerfCountersCollection::remove(PerfCounters*)+0x27) [0x6a0877]
163 5: (LevelDBStore::~LevelDBStore()+0x1b) [0x582b2b]
164 6: (LevelDBStore::~LevelDBStore()+0x9) [0x582da9]
165 7: (main()+0x1386) [0x48db16]
166 8: (__libc_start_main()+0xed) [0x658076d]
167 9: ceph-mon() [0x4909ad]
169 Signed-off-by: Sage Weil <sage@inktank.com>
170 (cherry picked from commit df2d06db6f3f7e858bdadcc8cd2b0ade432df413)
172 commit d284eaf9ce7d2022ba62562236d5fa41c26c1eb3
173 Author: Sage Weil <sage@inktank.com>
174 Date: Thu May 30 11:07:06 2013 -0700
176 mon: fix leak of health_monitor and config_key_service
178 Switch to using regular pointers here. The lifecycle of these services is
179 very simple such that refcounting is overkill.
181 Signed-off-by: Sage Weil <sage@inktank.com>
182 (cherry picked from commit c888d1d3f1b77e62d1a8796992e918d12a009b9d)
184 commit cc88cdfe8de02da675f9051e95b70da11f7fbe9e
185 Author: Sage Weil <sage@inktank.com>
186 Date: Wed May 29 17:54:17 2013 -0700
188 mon: return instead of exit(3) via preforker
190 This lets us run all the locally-scoped dtors so that leak checking will
193 Signed-off-by: Sage Weil <sage@inktank.com>
194 (cherry picked from commit 3c5706163b72245768958155d767abf561e6d96d)
196 commit 85d629a5f8e1deef9a562d9a7b371707d90c5ba1
197 Author: Sage Weil <sage@inktank.com>
198 Date: Thu May 30 14:57:42 2013 -0700
200 os/LevelDBStore: add perfcounters
202 Signed-off-by: Sage Weil <sage@inktank.com>
203 (cherry picked from commit 7802292e0a49be607d7ba139b44d5ea1f98e07e6)
205 commit 0109fa8ae70671c2d8ca19bcc95662d5f41cac66
206 Author: Sage Weil <sage@inktank.com>
207 Date: Thu May 30 14:36:41 2013 -0700
209 mon: make compaction bounds overlap
211 When we trim items N to M, compact over range (N-1) to M so that the
212 items in the queue will share bounds and get merged. There is no harm in
213 compacting over a larger range here when the lower bound is a key that
214 doesn't exist anyway.
216 Signed-off-by: Sage Weil <sage@inktank.com>
217 (cherry picked from commit a47ca583980523ee0108774b466718b303bd3f46)
219 commit 83b1edac07dd74b91ba2cdfe8b63236d7930c9b1
220 Author: Sage Weil <sage@inktank.com>
221 Date: Thu May 30 14:26:42 2013 -0700
223 os/LevelDBStore: merge adjacent ranges in compactionqueue
225 If we get behind and multiple adjacent ranges end up in the queue, merge
226 them so that we fire off compaction on larger ranges.
228 Signed-off-by: Sage Weil <sage@inktank.com>
229 (cherry picked from commit f628dd0e4a5ace079568773edfab29d9f764d4f0)
231 commit d14665e550d9b2dfc47684b73427042b0744127f
232 Author: Sage Weil <sage@inktank.com>
233 Date: Wed May 29 08:40:32 2013 -0700
235 mon: compact trimmed range, not entire prefix
237 This will reduce the work that leveldb is asked to do by only triggering
238 compaction of the keys that were just trimmed.
240 We ma want to further reduce the work by compacting less frequently, but
241 this is at least a step in that direction.
243 Signed-off-by: Sage Weil <sage@inktank.com>
244 (cherry picked from commit 6da4b20ca53fc8161485c8a99a6b333e23ace30e)
246 commit dcd9b793fb0b05976b55be029315114d6f1df0e5
247 Author: Sage Weil <sage@inktank.com>
248 Date: Wed May 29 08:35:44 2013 -0700
250 mon/MonitorDBStore: allow compaction of ranges
252 Allow a transaction to describe the compaction of a range of keys. Do this
253 in a backward compatible say, such that older code will interpret the
254 compaction of a prefix + range as compaction of the entire prefix. This
255 allows us to avoid introducing any new feature bits.
257 Signed-off-by: Sage Weil <sage@inktank.com>
258 (cherry picked from commit ab09f1e5c1305a64482ebbb5a6156a0bb12a63a4)
262 src/mon/MonitorDBStore.h
264 commit 6543da740a12c6ad085b807c9038d5b7b5aeaba6
265 Author: Sage Weil <sage@inktank.com>
266 Date: Wed May 29 08:34:13 2013 -0700
268 os/LevelDBStore: allow compaction of key ranges
270 Signed-off-by: Sage Weil <sage@inktank.com>
271 (cherry picked from commit e20c9a3f79ccfeb816ed634ca25de29fc5975ea8)
273 commit 87dcba2dd12a673fdc63ad64fb23e6e9f841d74f
274 Author: Sage Weil <sage@inktank.com>
275 Date: Tue May 28 16:35:55 2013 -0700
277 os/LevelDBStore: do compact_prefix() work asynchronously
279 We generally do not want to block while compacting a range of leveldb.
280 Push the blocking+waiting off to a separate thread. (leveldb will do what
281 it can to avoid blocking internally; no reason for us to wait explicitly.)
283 This addresses part of #5176.
285 Signed-off-by: Sage Weil <sage@inktank.com>
286 (cherry picked from commit 4af917d4478ec07734a69447420280880d775fa2)
288 commit bac5720b2a583e799c6961c733c4a9132a002440
289 Author: Sage Weil <sage@inktank.com>
290 Date: Sat May 11 17:36:13 2013 -0700
292 qa: rsync test: exclude /usr/local
294 Some plana have non-world-readable crap in /usr/local/samba. Avoid
295 /usr/local entirely for that and any similar landmines.
297 Signed-off-by: Sage Weil <sage@inktank.com>
298 (cherry picked from commit 82211f2197241c4f3d3135fd5d7f0aa776eaeeb6)
300 commit 9f1afe1a8e7906bdc54158a4813f011933f6a78f
301 Author: Sage Weil <sage@inktank.com>
302 Date: Fri May 31 21:16:54 2013 -0700
304 mon: fix uninitialized fields in MMonHealth
307 Signed-off-by: Sage Weil <sage@inktank.com>
308 (cherry picked from commit d7e2ab1451e284cd4273cca47eec75e1d323f113)
310 commit d5b67d49c87b4fe6a90cdd0bf4fac4f9d9377e81
311 Author: Samuel Just <sam.just@inktank.com>
312 Date: Fri May 31 13:44:39 2013 -0700
314 PGLog: only add entry to caller_ops in add() if reqid_is_indexed()
317 Signed-off-by: Samuel Just <sam.just@inktank.com>
318 Reviewed-by: Sage Weil <sage@inktank.com>
320 commit a5f6d8967fbbe87f34b23cfcceea84317b647f62
321 Author: Samuel Just <sam.just@inktank.com>
322 Date: Mon Apr 15 16:33:48 2013 -0700
324 PG: don't write out pg map epoch every handle_activate_map
326 We don't actually need to write out the pg map epoch on every
327 activate_map as long as:
328 a) the osd does not trim past the oldest pg map persisted
329 b) the pg does update the persisted map epoch from time
332 To that end, we now keep a reference to the last map persisted.
333 The OSD already does not trim past the oldest live OSDMapRef.
334 Second, handle_activate_map will trim if the difference between
335 the current map and the last_persisted_map is large enough.
338 Signed-off-by: Samuel Just <sam.just@inktank.com>
339 Reviewed-by: Greg Farnum <greg@inktank.com>
340 (cherry picked from commit 2c5a9f0e178843e7ed514708bab137def840ab89)
344 src/common/config_opts.h
346 - last_persisted_osdmap_ref gets set in the non-static
349 commit 9aaff1492a00e137f84af9b03e3a4608cea4b520
350 Author: Alexandre Marangone <alexandre.marangone@inktank.com>
351 Date: Fri May 31 12:33:11 2013 -0700
353 upstart: handle upper case in cluster name and id
355 Signed-off-by: Alexandre Marangone <alexandre.marangone@inktank.com>
356 (cherry picked from commit 851619ab6645967e5d7659d9b0eea63d5c402b15)
358 commit f87a19d34f9a03493eaca654dd176992676c5812
359 Author: Samuel Just <sam.just@inktank.com>
360 Date: Tue May 21 15:22:56 2013 -0700
362 OSDMonitor: skip new pools in update_pools_status() and get_pools_health()
364 New pools won't be full. mon->pgmon()->pg_map.pg_pool_sum[poolid] will
365 implicitly create an entry for poolid causing register_new_pgs() to assume that
366 the newly created pgs in the new pool are in fact a result of a split
367 preventing MOSDPGCreate messages from being sent out.
371 Signed-off-by: Samuel Just <sam.just@inktank.com>
372 Reviewed-by: Sage Weil <sage@inktank.com>
373 (cherry picked from commit 0289c445be0269157fa46bbf187c92639a13db46)
375 commit 34733bdc3b9da13195dfac8a25f98c6de64070d7
376 Author: Yehuda Sadeh <yehuda@inktank.com>
377 Date: Thu May 30 12:58:11 2013 -0700
379 rgw: only append prefetched data if reading from head
382 Backport: bobtail, cuttlefish
383 If the head object wrongfully contains data, but according to the
384 manifest we don't read from the head, we shouldn't copy the prefetched
385 data. Also fix the length calculation for that data.
387 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
388 Reviewed-by: Greg Farnum <greg@inktank.com>
389 (cherry picked from commit c5fc52ae0fc851444226abd54a202af227d7cf17)
391 commit 3d91301e5d788f9f9f52ba4067006f39a89e1531
392 Author: Yehuda Sadeh <yehuda@inktank.com>
393 Date: Thu May 30 09:34:21 2013 -0700
395 rgw: don't copy object idtag when copying object
398 When copying object we ended up also copying the original
399 object idtag which overrode the newly generated one. When
400 refcount put is called with the wrong idtag the count
403 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
404 Reviewed-by: Greg Farnum <greg@inktank.com>
405 (cherry picked from commit b1312f94edc016e604f1d05ccfe2c788677f51d1)
407 commit 131dca7d10d5b0b7dca5c5cbe9b0b54938805015
408 Author: Sage Weil <sage@inktank.com>
409 Date: Thu May 30 08:53:22 2013 -0700
411 debian: sync up postinst and prerm with latest
413 - do not use invoke-rc.d for upstart
414 - do not stop daemons on upgrade
415 - misc other cleanups
417 This corresponds to the state of master as of cf9aa7a.
419 Signed-off-by: Sage Weil <sage@inktank.com>
421 commit 8b7ca687ded06fe0b67d98e81fa1dabbed440853
422 Author: Joao Eduardo Luis <joao.luis@inktank.com>
423 Date: Thu May 30 18:17:28 2013 +0100
425 mon: Monitor: backup monmap using all ceph features instead of quorum's
427 When a monitor is freshly created and for some reason its initial sync is
428 aborted, it will end up with an incorrect backup monmap. This monmap is
429 incorrect in the sense that it will not contain the monitor's names as
430 it will expect on the next run.
432 This results from us being using the quorum features to encode the monmap
433 when backing it up, instead of CEPH_FEATURES_ALL.
437 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
438 (cherry picked from commit 626de387e617db457d6d431c16327c275b0e8a34)
440 commit 02ef6e918e4fe0368b02fbc700a4d921ae298dc3
441 Author: Sage Weil <sage@inktank.com>
442 Date: Wed May 29 09:49:11 2013 -0700
444 osd: do not assume head obc object exists when getting snapdir
446 For a list-snaps operation on the snapdir, do not assume that the obc for the
447 head means the object exists. This fixes a race between a head deletion and
448 a list-snaps that wrongly returns ENOENT, triggered by the DiffItersateStress
449 test when thrashing OSDs.
453 Signed-off-by: Sage Weil <sage@inktank.com>
454 Reviewed-by: Samuel Just <sam.just@inktank.com>
455 (cherry picked from commit 29e4e7e316fe3f3028e6930bb5987cfe3a5e59ab)
457 commit 85ad65e294f2b3d4bd1cfef6ae613e31d1cea635
458 Author: Sage Weil <sage@inktank.com>
459 Date: Wed May 29 16:50:04 2013 -0700
461 osd: initialize new_state field when we use it
463 If we use operator[] on a new int field its value is undefined; avoid
464 reading it or using |= et al until we initialize it.
467 Backport: cuttlefish, bobtail
468 Signed-off-by: Sage Weil <sage@inktank.com>
469 Reviewed-by: David Zafman <david.zafman@inktank.com>
470 (cherry picked from commit 50ac8917f175d1b107c18ecb025af1a7b103d634)
472 commit 65236a4a2541799a0d946df7f3048d4404678f64
473 Author: Samuel Just <sam.just@inktank.com>
474 Date: Tue May 28 11:10:05 2013 -0700
476 HashIndex: sync top directory during start_split,merge,col_split
478 Otherwise, the links might be ordered after the in progress
479 operation tag write. We need the in progress operation tag to
480 correctly recover from an interrupted merge, split, or col_split.
483 Backport: cuttlefish, bobtail
484 Signed-off-by: Samuel Just <sam.just@inktank.com>
485 Reviewed-by: Sage Weil <sage@inktank.com>
486 (cherry picked from commit 5bca9c38ef5187c7a97916970a7fa73b342755ac)
488 commit 0e127dc4ef16d19d56a3603ca90fa6b076d905b0
489 Author: Joao Eduardo Luis <joao.luis@inktank.com>
490 Date: Wed May 22 13:59:08 2013 +0100
492 mon: Paxos: get rid of the 'prepare_bootstrap()' mechanism
494 We don't need it after all. If we are in the middle of some proposal,
495 then we guarantee that said proposal is likely to be retried. If we
496 haven't yet proposed, then it's forever more likely that a client will
497 eventually retry the message that triggered this proposal.
499 Basically, this mechanism attempted at fixing a non-problem, and was in
500 fact triggering some unforeseen issues that would have required increasing
501 the code complexity for no good reason.
505 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
506 (cherry picked from commit e15d29094503f279d444eda246fc45c09f5535c9)
508 commit f29206bdd950d1e916a1422b4826caaf4a8cef3a
509 Author: Joao Eduardo Luis <joao.luis@inktank.com>
510 Date: Wed May 22 13:51:13 2013 +0100
512 mon: Paxos: finish queued proposals instead of clearing the list
514 By finishing these Contexts, we make sure the Contexts they enclose (to be
515 called once the proposal goes through) will behave as their were initially
516 planned: for instance, a C_Command() may retry the command if a -EAGAIN
517 is passed to 'finish_contexts', while a C_Trimmed() will simply set
518 'going_to_trim' to false.
520 This aims at fixing at least a bug in which Paxos will stop trimming if an
521 election is triggered while a trim is queued but not yet finished. Such
522 happens because it is the C_Trimmed() context that is responsible for
523 resetting 'going_to_trim' back to false. By clearing all the contexts on
524 the proposal list instead of finishing them, we stay forever unable to
525 trim Paxos again as 'going_to_trim' will stay True till the end of time as
530 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
531 (cherry picked from commit 586e8c2075f721456fbd40f738dab8ccfa657aa8)
533 commit b73d7c6b7e83fe8ce29e7150f25a4cca7647cccd
534 Author: Joao Eduardo Luis <joao.luis@inktank.com>
535 Date: Fri May 17 18:23:36 2013 +0100
537 mon: Paxos: finish_proposal() when we're finished recovering
539 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
540 (cherry picked from commit 2ff23fe784245f3b86bc98e0434b21a5318e0a7b)
542 commit 79eb61c4eadd9d94b3b8087d85b7210f4ab71d54
543 Merge: a460e53 8682248
544 Author: Samuel Just <sam.just@inktank.com>
545 Date: Thu May 23 20:09:29 2013 -0700
547 Merge branch 'wip_scrub_tphandle' into cuttlefish
550 Reviewed-by: Sage Weil <sage@inktank.com>
552 commit 86822485e518d61d7b2c02a6ff25eb2c4b4bc307
553 Author: Samuel Just <sam.just@inktank.com>
554 Date: Thu May 23 17:40:44 2013 -0700
556 PG: ping tphandle during omap loop as well
558 Signed-off-by: Samuel Just <sam.just@inktank.com>
560 commit d62716dd4ceb29032759bf84e864d214fe38a17c
561 Author: Samuel Just <sam.just@inktank.com>
562 Date: Thu May 23 15:24:39 2013 -0700
564 PG: reset timeout in _scan_list for each object, read chunk
566 Signed-off-by: Samuel Just <sam.just@inktank.com>
568 commit b8a25e08a638c31b9cfc2c1bf6d9bad40e921a9f
569 Author: Samuel Just <sam.just@inktank.com>
570 Date: Thu May 23 15:23:05 2013 -0700
572 OSD,PG: pass tphandle down to _scan_list
574 Signed-off-by: Samuel Just <sam.just@inktank.com>
576 commit a460e53ecac03e9c8f54c402a790e6d8cf75b38c
577 Author: Yehuda Sadeh <yehuda@inktank.com>
578 Date: Wed May 22 21:34:52 2013 -0700
580 rgw: iterate usage entries from correct entry
583 When iterating through usage entries, and when user id was
584 provided, we started at the user's first entry and not from
585 the entry indexed by the request start time.
586 This commit fixes the issue.
590 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
591 Reviewed-by: Greg Farnum <greg@inktank.com>
592 (cherry picked from commit 8b3a04dec8be13559716667d4b16cde9e9543feb)
594 commit 48e1b11bcbdbdf8661cb560c8563f91738034287
595 Author: Sage Weil <sage@inktank.com>
596 Date: Thu May 16 20:37:05 2013 -0700
598 sysvinit: fix enumeration of local daemons when specifying type only
600 - prepend $local to the $allconf list at the top
601 - remove $local special case for all case
602 - fix the type prefix checks to explicitly check for prefixes
604 Fugly bash, but works!
606 Backport: cuttlefish, bobtail
607 Signed-off-by: Sage Weil <sage@inktank.com>
608 Reviewed-by: Dan Mick <dan.mick@inktank.com>
609 (cherry picked from commit c80c6a032c8112eab4f80a01ea18e1fa2c7aa6ed)
611 commit 6e5f0db75a2bf591bd1fc0a5c2a93051f83ebc5d
612 Author: Sage Weil <sage@inktank.com>
613 Date: Wed May 22 09:47:29 2013 -0700
615 sysvinit: fix osd weight calculation on remote hosts
617 We need to do df on the remote host, not locally.
619 Simlarly, the ceph command uses the osd key, which exists remotely; run it there.
621 Signed-off-by: Sage Weil <sage@inktank.com>
622 (cherry picked from commit d81d0ea5c442699570bd93a90bea0d97a288a1e9)
624 commit 674e75bc0783cd9befe9c20e36fbc2cfdac62e5a
625 Author: Sage Weil <sage@inktank.com>
626 Date: Wed May 22 09:47:03 2013 -0700
628 sysvinit: use known hostname $host instead of (incorrectly) recalculating
630 We would need to do hostname -s on the remote node, not the local one.
631 But we already have $host; use it!
633 Reported-by: Xiaoxi Chen <xiaoxi.chen@intel.com>
634 Signed-off-by: Sage Weil <sage@inktank.com>
635 (cherry picked from commit caa15a34cb5d918c0c8b052cd012ec8a12fca150)
637 commit f4cd61d339419738355a26d7f4fce04eac9dedee
638 Author: Sage Weil <sage@inktank.com>
639 Date: Mon May 20 12:41:30 2013 -0700
641 mon: be a bit more verbose about osd mark down events
643 Put these in the cluster log; they are interesting.
645 Signed-off-by: Sage Weil <sage@inktank.com>
646 (cherry picked from commit 87767fb1fb9a52d11b11f0b641cebbd9998f089e)
648 commit e04b1894ed7f3a6e95007e58dae5b35357e5c507
649 Author: Samuel Just <sam.just@inktank.com>
650 Date: Mon May 13 14:23:00 2013 -0700
652 PG: subset_last_update must be at least log.tail
655 Backport: bobtail, cuttlefish
656 Signed-off-by: Samuel Just <sam.just@inktank.com>
657 Reviewed-by: David Zafman <david.zafman@inktank.com>
658 (cherry picked from commit 72bf5f4813c273210b5ced7f7793bc1bf813690c)
660 commit 3a02e00d1d3b904b312db283faedf4dff37b0f6f
661 Author: Samuel Just <sam.just@inktank.com>
662 Date: Tue May 14 16:35:48 2013 -0700
664 FileJournal: adjust write_pos prior to unlocking write_lock
666 In committed_thru, we use write_pos to reset the header.start value in cases
667 where seq is past the end of our journalq. It is therefore important that the
668 journalq be updated atomically with write_pos (that is, under the write_lock).
670 The call to align_bl() is moved into do_write in order to ensure that write_pos
671 is adjusted correctly prior to write_bl().
673 Also, we adjust pos at the end of write_bl() such that pos \in [get_top(),
674 header.max_size) after write_bl().
677 Signed-off-by: Samuel Just <sam.just@inktank.com>
678 Reviewed-by: Sage Weil <sage@inktank.com>
679 (cherry picked from commit eaf3abf3f9a7b13b81736aa558c9084a8f07fdbe)
681 commit 8a030eb0e1b61121f7b1e2b7a551bb213d1c428b
682 Author: Sage Weil <sage@inktank.com>
683 Date: Tue May 21 14:36:11 2013 -0700
685 mon: implement --extract-monmap <filename>
687 This will make for a simpler process for
688 http://docs.ceph.com/docs/master/rados/operations/add-or-rm-mons/#removing-monitors-from-an-unhealthy-cluster
690 Signed-off-by: Sage Weil <sage@inktank.com>
691 (cherry picked from commit c0268e27497a4d8228ef54da9d4ca12f3ac1f1bf)
693 commit 4ad13c945fd29a2d183f7ea6c6ac8a51d24dabe1
694 Author: Josh Durgin <josh.durgin@inktank.com>
695 Date: Thu May 16 15:28:40 2013 -0700
697 librbd: make image creation defaults configurable
699 Programs using older versions of the image creation functions can't
700 set newer parameters like image format and fancier striping.
702 Setting these options lets them use all the new functionality without
703 being patched and recompiled to use e.g. rbd_create3().
704 This is particularly useful for things like qemu-img, which does not
705 know how to create format 2 images yet.
708 backport: cuttlefish, bobtail
709 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
710 (cherry picked from commit aacc9adc4e9ca90bbe73ac153cc754a3a5b2c0a1)
712 commit 684444f88f2a7cf28f2e685c18f0771730a1d48f
713 Author: Josh Durgin <josh.durgin@inktank.com>
714 Date: Thu May 16 15:21:24 2013 -0700
716 rbd.py: fix stripe_unit() and stripe_count()
718 These matched older versions of the functions, but would segfault
719 using the current versions.
721 backport: cuttlefish, bobtail
722 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
723 (cherry picked from commit 53ee6f965e8f06c7256848210ad3c4f89d0cb5a0)
725 commit 9c7faf957fffb2721ccb915b68ca90ffb0d04a9f
726 Author: Josh Durgin <josh.durgin@inktank.com>
727 Date: Thu May 16 15:19:46 2013 -0700
729 cls_rbd: make sure stripe_unit is not larger than object size
731 Test a few other cases too.
733 backport: cuttlefish, bobtail
734 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
735 (cherry picked from commit 810306a2a76eec1c232fd28ec9c351e827fa3031)
737 commit 4071d7a79585ee2768d0a63819d99405a083369f
738 Author: Yehuda Sadeh <yehuda@inktank.com>
739 Date: Fri May 3 12:57:00 2013 -0700
741 rgw: protect ops log socket formatter
744 Ops log (through the unix domain socket) uses a formatter, which wasn't
747 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
748 (cherry picked from commit d48f1edb07a4d8727ac956f70e663c1b4e33e1dd)
750 commit 57a6a54db4020debd826a493cccca4695934f693
751 Author: Sage Weil <sage@inktank.com>
752 Date: Wed May 15 23:02:10 2013 -0700
754 Makefle: force char to be signed
756 On an armv7l build, we see errors like
758 warning: rgw/rgw_common.cc:626:16: comparison is always false due to limited range of data type [-Wtype-limits]
762 char c1 = hex_to_num(*src++);
766 Force char to be signed (regardless of any weird architecture's default)
767 to avoid risk of this leading to misbehavior.
769 Signed-off-by: Sage Weil <sage@inktank.com>
770 Reviewed-by: Dan Mick <dan.mick@inktank.com>
771 (cherry picked from commit 769a16d6674122f3b537f03e17514ad974bf2a2f)
773 commit 85fb422a084785176af3b694882964841e02195d
774 Author: Sage Weil <sage@inktank.com>
775 Date: Mon May 20 13:34:27 2013 -0700
777 debian: stop sysvinit on ceph.prerm
779 Signed-off-by: Sage Weil <sage@inktank.com>
780 (cherry picked from commit 2f193fb931ed09d921e6fa5a985ab87aa4874589)
782 commit eac3cd2d8334db7c70cbba433610136ff76e447f
783 Author: Mike Kelly <pioto@pioto.org>
784 Date: Thu May 16 12:29:50 2013 -0400
786 ceph df: fix si units for 'global' stats
788 si_t expects bytes, but it was being given kilobytes.
790 Signed-off-by: Mike Kelly <pioto@pioto.org>
791 (cherry picked from commit 0c2b738d8d07994fee4c73dd076ac9364a64bdb2)
793 commit 7bc7c9d4bc6f5bbadbb5e6f9844a0e1f66373c90
794 Author: Sage Weil <sage@inktank.com>
795 Date: Thu May 16 18:40:29 2013 -0700
797 udev: install disk/by-partuuid rules
799 Wheezy's udev (175-7.2) has broken rules for the /dev/disk/by-partuuid/
800 symlinks that ceph-disk relies on. Install parallel rules that work. On
801 new udev, this is harmless; old older udev, this will make life better.
805 Signed-off-by: Sage Weil <sage@inktank.com>
806 (cherry picked from commit d8d7113c35b59902902d487738888567e3a6b933)
808 commit c2acecbdab93177227a80a114ecc4ce71dbdbe49
809 Author: Sage Weil <sage@inktank.com>
810 Date: Thu May 16 13:17:45 2013 -0700
812 debian: make radosgw require matching version of librados2
814 ...indirectly via ceph-common. We get bad behavior when they diverge, I
815 think because of libcommon.la being linked both statically and dynamically.
818 Backport: cuttlefish, bobtail
819 Signed-off-by: Sage Weil <sage@inktank.com>
820 Reviewed-by: Gary Lowell <gary.lowell@inktank.com>
821 (cherry picked from commit 604c83ff18f9a40c4f44bc8483ef22ff41efc8ad)
823 commit 4c0d3eb72c045c39e8b38fd5abf9ddfbb2219f9c
824 Author: Sage Weil <sage@inktank.com>
825 Date: Fri May 10 22:14:05 2013 -0700
827 mon: fix validatation of mds ids in mon commands
830 Signed-off-by: Sage Weil <sage@inktank.com>
831 (cherry picked from commit 5c305d63043762027323052b4bb3ae3063665c6f)