1 commit 63b0f854d1cef490624de5d6cf9039735c7de5ca
2 Author: Gary Lowell <gary.lowell@inktank.com>
3 Date: Mon Mar 25 11:02:31 2013 -0700
7 commit 90ab120a3c08f30654207c9c67fb0a4ff1476aff
8 Author: Yehuda Sadeh <yehuda@inktank.com>
9 Date: Mon Mar 25 09:50:33 2013 -0700
11 rgw: bucket index ops on system buckets shouldn't do anything
15 On certain bucket index operations we didn't check whether
16 the bucket was a system bucket, which caused the operations
17 to fail. This triggered an error message on bucket removal
20 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
21 Reviewed-by: Greg Farnum <greg@inktank.com>
22 (cherry picked from commit 70e0ee8ba955322832f0c366537ddf7a0288761e)
24 commit b8657fa08d0fe176c71a60c1654c53c3156a9c53
25 Author: Josh Durgin <josh.durgin@inktank.com>
26 Date: Mon Feb 25 15:02:50 2013 -0800
28 systest: restrict list error acceptance
30 Only ignore errors after the midway point if the midway_sem_post is
33 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
34 (cherry picked from commit 5b24a68b6e7d57bac688021b822fb2f73494c3e9)
36 commit 5af3d64f62b30cbaac13ccc7c23fc884ee1b7ae3
37 Author: Josh Durgin <josh.durgin@inktank.com>
38 Date: Mon Feb 25 14:55:34 2013 -0800
40 systest: fix race with pool deletion
42 The second test have pool deletion and object listing wait on the same
43 semaphore to connect and start. This led to errors sometimes when the
44 pool was deleted before it could be opened by the listing process. Add
45 another semaphore so the pool deletion happens only after the listing
49 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
50 (cherry picked from commit b0271e390564119e998e18189282252d54f75eb6)
52 commit 3569489b541ac0643520d20b08c788c26dfaff7f
53 Author: Sage Weil <sage@inktank.com>
54 Date: Tue Mar 19 14:26:16 2013 -0700
56 os/FileJournal: fix aio self-throttling deadlock
58 This block of code tries to limit the number of aios in flight by waiting
59 for the amount of data to be written to grow relative to a function of the
60 number of aios. Strictly speaking, the condition we are waiting for is a
61 function of both aio_num and the write queue, but we are only woken by
62 changes in aio_num, and were (in rare cases) waiting when aio_num == 0 and
63 there was no possibility of being woken.
65 Fix this by verifying that aio_num > 0, and restructuring the loop to
66 recheck that condition on each wakeup.
69 Signed-off-by: Sage Weil <sage@inktank.com>
70 Reviewed-by: Samuel Just <sam.just@inktank.com>
71 (cherry picked from commit e5940da9a534821d0d8f872c13f9ac26fb05a0f5)
73 commit 664ffa7d2178e486b00fa0706067f19b1bb9ab82
74 Author: Sage Weil <sage@inktank.com>
75 Date: Fri Mar 22 13:25:49 2013 -0700
77 common/MemoryModel: remove logging to /tmp/memlog
79 This was a hack for dev purposes ages ago; remove it. The predictable
80 filename is a security issue.
84 Reported-by: Michael Scherer <misc@zarb.org>
85 Signed-off-by: Sage Weil <sage@inktank.com>
86 Reviewed-by: Dan Mick <dan.mick@inktank.com>
87 (cherry picked from commit c524e2e01da41ab5b6362c117939ea1efbd98095)
89 commit 901fa435bb4be29b72cec39e5ff77570d13cbc6c
90 Author: Sage Weil <sage@inktank.com>
91 Date: Fri Mar 22 13:25:43 2013 -0700
93 init-ceph: clean up temp ceph.conf filename on exit
95 Signed-off-by: Sage Weil <sage@inktank.com>
96 Reviewed-by: Dan Mick <dan.mick@inktank.com>
97 (cherry picked from commit 6a7ad2eac1db6abca3d7edb23ca9b80751400a23)
99 commit 951f7e562afb4ae91b8d037ef793a96779461e96
100 Author: Sage Weil <sage@inktank.com>
101 Date: Fri Mar 22 13:25:33 2013 -0700
103 init-ceph: push temp conf file to a unique location on remote host
105 The predictable file name is a security problem.
109 Reported-by: Michael Scherer <misc@zarb.org>
110 Signed-off-by: Sage Weil <sage@inktank.com>
111 Reviewed-by: Dan Mick <dan.mick@inktank.com>
112 (cherry picked from commit 051734522fea92878dd8139f28ec4e6b01371ede)
114 commit dd670afccfd3ae58c03ea3b4c213dd8db22f618e
115 Author: Sage Weil <sage@inktank.com>
116 Date: Fri Mar 22 13:25:23 2013 -0700
118 mkcephfs: make remote temp directory name unique
120 The predictable file name is a security problem.
124 Reported-by: Michael Scherer <misc@zarb.org>
125 Signed-off-by: Sage Weil <sage@inktank.com>
126 Reviewed-by: Dan Mick <dan.mick@inktank.com>
127 (cherry picked from commit f463ef78d77b11b5ad78b31e9a3a88d0a6e62bca)
129 commit 1cd51eb63a91fa0aa395c41572cfee4d53fbc7f5
130 Author: Samuel Just <sam.just@inktank.com>
131 Date: Fri Mar 22 13:51:14 2013 -0700
133 PG::GetMissing: need to check need_up_thru in MLogRec handler
137 Signed-off-by: Samuel Just <sam.just@inktank.com>
138 Reviewed-by: Sage Weil <sage@inktank.com>
139 (cherry picked from commit 4fe4deafbe1758a6b3570048aca57485bd562440)
141 commit d866814285667c7f9cd9205dca6cf95f3634c5fc
142 Author: Samuel Just <sam.just@inktank.com>
143 Date: Fri Mar 22 13:48:49 2013 -0700
145 PG,osd_types: improve check_new_interval debugging
148 Signed-off-by: Samuel Just <sam.just@inktank.com>
149 Reviewed-by: Sage Weil <sage@inktank.com>
150 (cherry picked from commit d611eba9caf45f2d977c627b123462a073f523a4)
152 commit fc3406f395103737e5784611e2b38fd1c4f26369
153 Author: Samuel Just <sam.just@inktank.com>
154 Date: Tue Mar 5 16:06:20 2013 -0800
156 FileStore: fix reversed collection_empty return value
160 Signed-off-by: Samuel Just <sam.just@inktank.com>
161 Reviewed-by: Greg Farnum <greg@inktank.com>
162 (cherry picked from commit 7a434d10da0f77e1b329de0b06b6645cd73cc81b)
167 commit 9f7c6ad395653b1039ea957cfaa7689699f510e7
168 Author: Samuel Just <sam.just@inktank.com>
169 Date: Mon Feb 11 12:52:07 2013 -0800
171 FileStore: set replay guard on create_collection
173 This should prevent sequences like:
180 from causing trouble by preventing the rmcoll
181 and mkcoll from being replayed.
185 Signed-off-by: Samuel Just <sam.just@inktank.com>
186 (cherry picked from commit 411770c45734c9827745ddc4018d86c14f2858a6)
188 commit 9932c9f0a0503f1902549a3c39e5d51488fff3ef
189 Author: Samuel Just <sam.just@inktank.com>
190 Date: Mon Feb 11 12:24:14 2013 -0800
192 FileStore: _split_collection should not create the collection
194 This will simplify adding a replay guard to create_collection.
197 Signed-off-by: Samuel Just <sam.just@inktank.com>
198 (cherry picked from commit b184ff581a08c9e6ce5b858f06ccbe9d0e2a170b)
200 commit 7e7ea8266ee45d96863e1edc5a567dcc2977680c
201 Author: Sage Weil <sage@inktank.com>
202 Date: Fri Feb 22 15:15:27 2013 -0800
204 client: use 4MB f_bsize and f_frsize for statfs
208 Block size: 1048576 Fundamental block size: 1048576
210 and the df(1) arithmetic works out. New stat(1) reports:
212 Block size: 1048576 Fundamental block size: 4096
214 which is what we are shoving into statvfs, but we have the b_size and
215 fr_size arithmetic swapped. However, doing the *correct* reporting would
216 then break the old stat by making both sizes appear to be 4KB (or
219 Sidestep the issue by making *both* values 4MB.. which is both large enough
220 to report large FS sizes, and also the default stripe size and thus a
221 "reasonable" value to report for a block size.
223 Perhaps in the future, when we no longer care about old userland, we can
224 report the page size for f_bsize, which is probably the "most correct"
227 Fixes: #3794. See also #3793.
228 Signed-off-by: Sage Weil <sage@inktank.com>
229 Reviewed-by: Greg Farnum <greg@inktank.com>
230 (cherry picked from commit 7c94083643891c9d66a117352f312b268bdb1135)
232 commit 66dca7730006dffbce12e6b27016ea5469b7691b
233 Author: Sage Weil <sage@inktank.com>
234 Date: Mon Feb 18 17:39:46 2013 -0800
236 os/FileStore: check replay guard on src for collection rename
238 This avoids a problematic sequence like:
247 - remove B/1...100 (fails but tolerated)
248 - destroy B/ (fails with ENOTEMPTY)
250 Signed-off-by: Sage Weil <sage@inktank.com>
251 Reviewed-by: Samuel Just <sam.just@inktank.com>
252 (cherry picked from commit 5fc83c8d9887d2a916af11436ccc94fcbfe59b7a)
254 commit ea570591b0d9dda102bece0f4cc7d263c3e2a3b0
255 Author: Samuel Just <sam.just@inktank.com>
256 Date: Fri Feb 22 14:12:28 2013 -0800
258 PG::proc_replica_log: oinfo.last_complete must be *before* first entry in omissing
261 Signed-off-by: Samuel Just <sam.just@inktank.com>
262 (cherry picked from commit 2dae6a68ee85a20220ee940dbe33a2144d43457b)
264 commit aa38d87c6237ccfb8b3b03bc7c28dcadb927237b
265 Author: Sage Weil <sage@inktank.com>
266 Date: Thu Feb 21 17:55:21 2013 -0800
268 osd/PG: fix typo, missing -> omissing
270 From ce7ffc34408bf32c66dc07e6f42d54b7ec489d41.
272 Signed-off-by: Sage Weil <sage@inktank.com>
273 (cherry picked from commit dc181224abf6fb8fc583730ae3d90acdf0b80f39)
275 commit ac914834ba4fb0e4770f63c60713be7a48aa3fa1
276 Author: Samuel Just <sam.just@inktank.com>
277 Date: Thu Feb 21 15:31:36 2013 -0800
279 PG::proc_replica_log: adjust oinfo.last_complete based on omissing
281 Otherwise, search_for_missing may neglect to check the missing
282 set for some objects assuming that if the need version is
283 prior to last_complete, the replica must have it.
286 Signed-off-by: Samuel Just <sam.just@inktank.com>
287 (cherry picked from commit ce7ffc34408bf32c66dc07e6f42d54b7ec489d41)
289 commit 7f39704b0903fb11e3f1e102599bf0c540f93623
290 Author: Sage Weil <sage@inktank.com>
291 Date: Sat Feb 9 00:05:33 2013 -0800
293 osd: fix load_pgs collection handling
295 On a _TEMP pg, is_pg() would succeed, which meant we weren't actually
296 hitting the cleanup checks. Instead, restructure this loop as positive
297 checks and handle each type of collection we understand.
299 This fixes _TEMP cleanup.
301 Signed-off-by: Sage Weil <sage@inktank.com>
302 Reviewed-by: Samuel Just <sam.just@inktank.com>
303 (cherry picked from commit b19b6dced85617d594c15631571202aab2f94ae8)
305 commit 0a09be99f3e9d2b68a052ff63ea858a7c8e5626c
306 Author: Sage Weil <sage@inktank.com>
307 Date: Sat Feb 9 00:04:29 2013 -0800
309 osd: fix load_pgs handling of pg dirs without a head
311 If there is a pgid that passes coll_t::is_pg() but there is no head, we
312 will populate the pgs map but then fail later when we try to do
313 read_state. This is a side-effect of 55f8579.
315 Take explicit note of _head collections we see, and then warn when we
316 find stray snap collections.
318 Signed-off-by: Sage Weil <sage@inktank.com>
319 Reviewed-by: Samuel Just <sam.just@inktank.com>
320 (cherry picked from commit 1f80a0b576c0af1931f743ad988b6293cbf2d6d9)
322 commit 2dc25dfd47de8d3602bdb3699c82b8b118b0fb28
323 Author: Samuel Just <sam.just@inktank.com>
324 Date: Thu Feb 7 13:34:47 2013 -0800
326 OSD::load_pgs: first scan colls before initing PGs
329 Signed-off-by: Samuel Just <sam.just@inktank.com>
330 Reviewed-by: Sage Weil <sage@inktank.com>
331 (cherry picked from commit 073f58ede2e473af91f76d01679631c169274af7)
333 commit 2fd8db29ccf56fd5473826f20e351edfe200419d
334 Author: David Zafman <david.zafman@inktank.com>
335 Date: Tue Jan 8 19:24:13 2013 -0800
337 osd: Add digest of omap for deep-scrub
339 Add ScrubMap encode/decode v4 message with omap digest
340 Compute digest of header and key/value. Use bufferlist
341 to reflect structure and compute as we go, clearing
342 bufferlist to reduce memory usage.
344 Signed-off-by: David Zafman <david.zafman@inktank.com>
345 Reviewed-by: Samuel Just <sam.just@inktank.com>
346 (cherry picked from commit 509a93e89f04d7e9393090563cf7be8e0ea53891)
348 commit ee59f517ccc195e591e0b1b6ddf6d1d054c4e2fd
349 Author: Samuel Just <sam.just@inktank.com>
350 Date: Fri Mar 15 15:13:46 2013 -0700
352 OSD: split temp collection as well
354 Otherwise, when we eventually remove the temp collection, there might be
355 objects in the temp collection which were independently pulled into the child
356 pg collection. Thus, removing the old stale parent link from its temp
357 collection also blasts the omap entries and snap mappings for the real child
362 Signed-off-by: Samuel Just <sam.just@inktank.com>
363 Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
364 (cherry picked from commit f8d66e87a5c155b027cc6249006b83b4ac9b6c9b)
366 commit 7ec68e2d45b431a3a77458939622e6ea1e396d0e
367 Author: Samuel Just <sam.just@inktank.com>
368 Date: Thu Mar 14 19:59:36 2013 -0700
370 PG: ignore non MISSING pg query in ReplicaActive
372 1) Replica sends notify
373 2) Prior to processing notify, primary queues query to replica
374 3) Primary processes notify and activates sending MOSDPGLog
376 4) Primary does do_notifies at end of process_peering_events
378 5) Replica sees MOSDPGLog and activates
379 6) Replica sees Query and asserts.
381 In the above case, the Replica should simply ignore the old
386 Signed-off-by: Samuel Just <sam.just@inktank.com>
387 Reviewed-by: Sage Weil <sage@inktank.com>
388 (cherry picked from commit 8222cbc8f35c359a35f8381ad90ff0eed5615dac)
390 commit b279b148b38c755d2dcca6524d7b8a95daf123a0
391 Author: Samuel Just <sam.just@inktank.com>
392 Date: Wed Mar 13 16:04:23 2013 -0700
394 FileJournal: queue_pos \in [get_top(), header.max_size)
396 If queue_pos == header.max_size when we create the entry
397 header magic, the entry will be rejected at get_top() on
402 Signed-off-by: Samuel Just <sam.just@inktank.com>
403 Reviewed-by: Sage Weil <sage@inktank.com>
404 (cherry picked from commit de8edb732e3a5ce4471670e43cfe6357ae6a2758)
406 commit 56e0a129b0d9ad463f4300999ed2c15173cbe603
407 Author: Samuel Just <sam.just@inktank.com>
408 Date: Thu Mar 14 18:52:02 2013 -0700
410 OSD: expand_pg_num after pg removes
413 1) expand_pg_num removes a splitting pg entry
414 2) peering thread grabs pg lock and starts split
415 3) OSD::consume_map grabs pg lock and starts removal
417 At step 2), we run afoul of the assert(is_splitting)
418 check in split_pgs. This way, the would be splitting
419 pg is marked as removed prior to the splitting state
424 Signed-off-by: Samuel Just <sam.just@inktank.com>
425 Reviewed-by: Sage Weil <sage@inktank.com>
426 (cherry picked from commit f1b031b3cf195cf6df3d3c47c7d606fba63ed4c4)
428 commit d6e678a62950ccc16d010a76500705e176ba9c73
429 Author: Sage Weil <sage@inktank.com>
430 Date: Mon Feb 11 06:23:54 2013 -0800
432 osd: update snap collections for sub_op_modify log records conditionaly
434 The only remaining caller is sub_op_modify(). If we do have a non-empty
435 op transaction, we want to do this update, regardless of what we think
436 last_backfill is (our notion may be not completely in sync with the
437 primary). In particular, our last_backfill may be the same object but
438 a different snapid, but the primary disagrees and is pushing an op
441 Instead, update the collections if we have a non-empty transaction.
443 Signed-off-by: Sage Weil <sage@inktank.com>
444 Reviewed-by: Samuel Just <sam.just@inktank.com>
445 (cherry picked from commit 31e911b63d326bdd06981ec4029ad71b7479ed70)
447 commit 51c2c4d7ccb25617967689d7e531eb145ee7011d
448 Author: Sage Weil <sage@inktank.com>
449 Date: Sun Feb 10 17:02:45 2013 -0800
451 osd: include snaps in pg_log_entry_t::dump()
453 Signed-off-by: Sage Weil <sage@inktank.com>
454 Reviewed-by: Samuel Just <sam.just@inktank.com>
455 (cherry picked from commit 715d8717a0e8a08fbe97a3e7d3ffd33aa9529d90)
457 commit f0bf68ff500f7337103aef2a9d6d10c3597e410f
458 Author: Sage Weil <sage@inktank.com>
459 Date: Sun Feb 10 16:59:48 2013 -0800
461 osd: unconditionally encode snaps buffer
463 Previously we would only encode the updated snaps vector for CLONE ops.
464 This doesn't work for MODIFY ops generated by the snap trimmer, which
465 may also adjust the clone collections. It is also possible that other
466 operations may need to populate this field in the future (e.g.,
467 LOST_REVERT may, although it currently does not).
469 Fixes: #4071, and possibly #4051.
471 Signed-off-by: Sage Weil <sage@inktank.com>
472 Reviewed-by: Samuel Just <sam.just@inktank.com>
473 (cherry picked from commit 54b6dd924fea3af982f3d729150b6449f318daf2)
475 commit aa74fabd90b776e9938f3d59ea5ed04bd4027dbb
476 Author: Sage Weil <sage@inktank.com>
477 Date: Sun Feb 10 10:57:12 2013 -0800
479 osd: improve debug output on snap collections
481 Signed-off-by: Sage Weil <sage@inktank.com>
482 Reviewed-by: Samuel Just <sam.just@inktank.com>
483 (cherry picked from commit 8b05492ca5f1479589bb19c1ce058b0d0988b74f)
485 commit a30903c6adaa023587d3147179d6038ad37ca520
486 Author: Samuel Just <sam.just@inktank.com>
487 Date: Thu Mar 7 12:53:51 2013 -0800
489 PG: check_recovery_sources must happen even if not active
491 missing_loc/missing_loc_sources also must be cleaned up
492 if a peer goes down during peering:
494 1) pg is in GetInfo, acting is [3,1]
495 2) we find object A on osd [0] in GetInfo
496 3) 0 goes down, no new peering interval since it is neither up nor
497 acting, but peer_missing[0] is removed.
498 4) pg goes active and try to pull A from 0 since missing_loc did not get
503 Signed-off-by: Samuel Just <sam.just@inktank.com>
504 Reviewed-by: Greg Farnum <greg@inktank.com>
505 (cherry picked from commit de22b186c497ce151217aecf17a8d35cdbf549bb)
507 commit 19f6df28337307ce6d8617524f88f222c0ffcab3
508 Author: Samuel Just <sam.just@inktank.com>
509 Date: Tue Mar 5 15:49:26 2013 -0800
511 HashIndex: _collection_list_partial must tolerate NULL next
515 Signed-off-by: Samuel Just <sam.just@inktank.com>
516 Reviewed-by: Greg Farnum <greg@inktank.com>
517 (cherry picked from commit ce4432adc67dc2fc06dd21ea08e59d179496bcc6)
519 commit 1b5c1564bad3574db257cb17d48190958d870f51
520 Author: Samuel Just <sam.just@inktank.com>
521 Date: Tue Mar 5 14:35:39 2013 -0800
523 OSD: lock not needed in ~DeletingState()
525 No further refs to the object can remain at this point.
526 Furthermore, the callbacks might lock mutexes of their
531 Signed-off-by: Samuel Just <sam.just@inktank.com>
532 Reviewed-by: Greg Farnum <greg@inktank.com>
533 (cherry picked from commit e4bf1bcab159d7c5b720f5da01877c0f67c16d16)
535 commit 50d2d4a77f339d489703c1125dc5c1159d08ca8a
536 Author: Samuel Just <sam.just@inktank.com>
537 Date: Sun Mar 10 12:50:01 2013 -0700
539 ReplicatedPG: don't leak reservation on removal
542 Signed-off-by: Samuel Just <sam.just@inktank.com>
543 Reviewed-by: Greg Farnum <greg@inktank.com>
544 (cherry picked from commit 32bf131e0141faf407b5ff993f75f97516b27c12)
548 src/osd/ReplicatedPG.cc
550 commit c6b7c4be3fc44d065c29ea76b744b39272c2ba05
551 Author: Yehuda Sadeh <yehuda@inktank.com>
552 Date: Tue Mar 12 12:56:01 2013 -0700
554 rgw: set up curl with CURL_NOSIGNAL
558 Apparently, libcurl needs that in order to be thread safe. Side
559 effect is that if libcurl is not compiled with c-ares support,
560 domain name lookups are not going to time out.
561 Issue affected keystone.
563 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
564 Reviewed-by: Greg Farnum <greg@inktank.com>
565 (cherry picked from commit 88725316ddcfa02ff110e659f7a8131dc1ea2cfc)
567 commit 8c16a376a9cbfd6812824fb71e2cc9935e14e667
568 Author: Sage Weil <sage@inktank.com>
569 Date: Fri Mar 8 08:56:44 2013 -0800
571 osd: mark down connections from old peers
573 Close out any connection with an old peer. This avoids a race like:
576 - we get map, mark down the con
577 - they reconnect and try to send us some stuff
578 - we share our map to tell them they are old and dead, but leave the con
581 - peer marks itself up a few times, eventually reuses the same port
582 - sends messages on their fresh con
583 - we discard because of our old con
585 This could cause a tight reconnect loop, but it is better than wrong
588 Other possible fixes:
589 - make addr nonce truly unique (augment pid in nonce)
590 - make a smarter 'disposable' msgr state (bleh)
592 Signed-off-by: Sage Weil <sage@inktank.com>
593 (cherry picked from commit 881e9d850c6762290f8be24da9e74b9dc112f1c9)
595 commit 1affc8b6e8cb82510aed524f90a6165c6edf7513
596 Author: Sage Weil <sage@inktank.com>
597 Date: Fri Mar 8 08:53:40 2013 -0800
599 osd/PG: rename require_same_or_newer_map -> is_same_or_newer_map
601 This avoids confusion with the OSD method of the same name, and better
602 matches what the function tests (and does not do).
604 Signed-off-by: Sage Weil <sage@inktank.com>
605 (cherry picked from commit ba7e815a18cad110525f228db1b3fe39e011409e)
609 src/osd/ReplicatedPG.cc
611 commit 7ca55d36c4007be5880102fa1ed314d3d66e9d87
612 Author: Sage Weil <sage@inktank.com>
613 Date: Mon Mar 11 16:25:16 2013 -0700
615 log: drop default 'log max recent' from 100k -> 10k
619 Signed-off-by: Sage Weil <sage@inktank.com>
620 (cherry picked from commit c021c5ccf0c063cccd7314964420405cea6406de)
622 commit 33ac786ddc14d0904f4835cbf684a51f4815d45b
623 Author: Jan Harkes <jaharkes@cs.cmu.edu>
624 Date: Fri Mar 8 12:45:57 2013 -0500
626 Fix radosgw actually reloading after rotating logs.
628 The --signal argument to Debian's start-stop-daemon doesn't
629 make it send a signal, but defines which signal should be send
630 when --stop is specified.
632 Signed-off-by: Jan Harkes <jaharkes@cs.cmu.edu>
633 (cherry picked from commit 44f1cc5bc42f9bb6d5a386037408d2de17dc5413)
635 commit 6838b8db9a3d4ceaf121a261e2ded4af9bfeaeb9
636 Author: Josh Durgin <josh.durgin@inktank.com>
637 Date: Wed Mar 6 17:42:03 2013 -0800
639 common: reduce default in-memory logs for non-daemons
641 The default of 100000 can result in hundreds of MBs of extra memory
642 used. This was most obvious when using librbd with caching enabled,
643 since there was a dout(0) accidentally left in the ObjectCacher.
647 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
648 (cherry picked from commit 7c208d2f8e3f28f4055a4ae51eceae892dcef1dc)
650 commit f80f64cf024bd7519d5a1fb2a5698db97a003ce8
651 Author: Sage Weil <sage@inktank.com>
652 Date: Fri Feb 22 17:01:53 2013 -0800
654 osd: allow (some) log trim when degraded, but not during recovery
656 We allow some trim during degraded, although we keep more entries around to
657 improve our chances of a restarting OSD of doing log-based recovery.
659 Still disallow during recovery...
661 Signed-off-by: Sage Weil <sage@inktank.com>
662 Reviewed-by: Samuel Just <sam.just@inktank.com>
663 (cherry picked from commit 6d89b34e5608c71b49ef33ab58340e90bd8da6e4)
665 commit e502a65da1779564059e74c09ab87aca1d901bd2
666 Author: Sage Weil <sage@inktank.com>
667 Date: Mon Feb 25 15:33:35 2013 -0800
669 osd: restructure calc_trim
671 No functional change, except that we log more debug, yay!
673 Signed-off-by: Sage Weil <sage@inktank.com>
674 Reviewed-by: Samuel Just <sam.just@inktank.com>
675 (cherry picked from commit 86df164d04f6e31a0f20bbb94dbce0599c0e8b3d)
677 commit 3a3123033d7d9e1574be3fb18c807eeb0f4678eb
678 Author: Sage Weil <sage@inktank.com>
679 Date: Fri Feb 22 16:48:02 2013 -0800
681 osd: allow pg log trim during (non-classic) scrub
683 Chunky (and deep) scrub do not care about PG log trimming. Classic scrub
686 Deep scrub can take a long time, so not trimming the log during that period
687 may eat lots of RAM; avoid that!
690 Signed-off-by: Sage Weil <sage@inktank.com>
691 Reviewed-by: Samuel Just <sam.just@inktank.com>
692 (cherry picked from commit 0ba8db6b664205348d5499937759916eac0997bf)
694 commit cde0f224dac21b2e17e06d396ac52e18034a4262
695 Author: Sage Weil <sage@inktank.com>
696 Date: Thu Feb 28 12:46:00 2013 -0800
698 msgr: drop messages on cons with CLOSED Pipes
700 Back in commit 6339c5d43974f4b495f15d199e01a141e74235f5, we tried to make
701 this deal with a race between a faulting pipe and new messages being
702 queued. The sequence is
704 - fault starts on pipe
705 - fault drops pipe_lock to unregister the pipe
706 - user (objecter) queues new message on the con
707 - submit_message reopens a Pipe (due to this bug)
708 - the message managed to make it out over the wire
709 - fault finishes faulting, calls ms_reset
710 - user (objecter) closes the con
711 - user (objecter) resends everything
713 It appears as though the previous patch *meant* to drop *m on the floor in
714 this case, which is what this patch does. And that fixes the crash I am
717 Signed-off-by: Sage Weil <sage@inktank.com>
718 Reviewed-by: Greg Farnum <greg@inktank.com>
719 (cherry picked from commit 0f42eddef5da6c1babe9ed51ceaa3212a42c2ec4)
721 commit 2fa2546c90e7a6eab9a3de7ab4e4271f8b25bad6
722 Author: Concubidated <tyler.brekke@inktank.com>
723 Date: Fri Mar 8 13:44:39 2013 -0800
725 Fix output of 'ceph osd tree --format=json'
727 Signed-off-by: Tyler Brekke <tyler.brekke@inktank.com>
728 (cherry picked from commit 9bcba944c6586ad5f007c0a30e69c6b5a886510b)
730 commit b70e2c270b9eb3fce673b7e51b527ebf88214f14
731 Author: Sam Lang <sam.lang@inktank.com>
732 Date: Tue Feb 12 11:32:29 2013 -0600
734 deb: Add ceph-coverage to ceph-test deb package
736 Teuthology uses the ceph-coverage script extensively
737 and expects it to be installed by the ceph task. Add
738 the script to the ceph-test debian package so that it
739 gets installed for that use case.
741 Signed-off-by: Sam Lang <sam.lang@inktank.com>
742 (cherry picked from commit 376cca2d4d4f548ce6b00b4fc2928d2e6d41038f)
744 commit ca9aac785eb9e2d1ee955792d2f4d1d911727fb3
745 Author: Yehuda Sadeh <yehuda@inktank.com>
746 Date: Fri Feb 22 15:04:37 2013 -0800
748 rgw: set attrs on various list bucket xml results (swift)
751 The list buckets operation was missing some attrs on the different
752 xml result entities. This fixes it.
754 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
755 (cherry picked from commit 4384e59ad046afc9ec53a2d2f1fff6a86e645505)
757 commit e39660c901756d5e722308e72a9d8ee4893f70f7
758 Author: Yehuda Sadeh <yehuda@inktank.com>
759 Date: Fri Feb 22 15:02:02 2013 -0800
761 formatter: add the ability to dump attrs in xml entities
763 xml entities may have attrs assigned to them. Add the ability
764 to set them. A usage example:
766 formatter->open_array_section_with_attrs("container",
767 FormatterAttrs("name", "foo", NULL));
769 This will generate the following xml entity:
770 <container name="foo">
772 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
773 Reviewed-by: Greg Farnum <greg@inktank.com>
774 (cherry picked from commit 7cb6ee28073824591d8132a87ea09a11c44efd66)
777 src/common/Formatter.cc
779 commit 0304b85e5c12c30459b94ec5d332dfaa04d20d5b
780 Author: Yehuda Sadeh <yehuda@inktank.com>
781 Date: Wed Mar 6 19:32:21 2013 -0800
783 rgw: don't iterate through all objects when in namespace
786 Backport: argonaut, bobtail
787 When listing objects in namespace don't iterate through all the
788 objects, only go though the ones that starts with the namespace
791 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
792 (cherry picked from commit 6669e73fa50e3908ec825ee030c31a6dbede6ac0)
794 commit d0dafaa77f96fa91785df0796806ea07fd93f319
795 Author: Josh Durgin <josh.durgin@inktank.com>
796 Date: Thu Feb 28 12:13:45 2013 -0800
798 ObjectCacher: fix debug log level in split
800 Level 0 should never be used for this kind of debugging.
802 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
803 Reviewed-by: Dan Mick <dan.mick@inktank.com>
804 (cherry picked from commit cb3ee33532fb60665f39f6ccb1d69d67279fd5e1)
806 commit ff014ed26bfa4fbf40acffec4eb4805b95a9982c
807 Author: Dan Mick <dmick@danceorelse.org>
808 Date: Thu Jan 24 13:38:25 2013 -0800
810 rados: remove unused "check_stdio" parameter
812 Signed-off-by: Dan Mick <dan.mick@inktank.com>
813 (cherry picked from commit bb860e49a7faeaf552538a9492ef0ba738c99760)
815 commit e89884da9f76b713372a79b772ba3cc2f3b03048
816 Author: Sage Weil <sage@inktank.com>
817 Date: Wed Jan 23 21:31:11 2013 -0800
819 rados: obey op_size for 'get'
821 Otherwise we try to read the whole object in one go, which doesn't bode
822 well for large objects (either non-optimal or simply broken).
824 Signed-off-by: Sage Weil <sage@inktank.com>
825 Reviewed-by: Dan Mick <dan.mick@inktank.com>
826 (cherry picked from commit 234becd3447a679a919af458440bc31c8bd6b84f)
828 commit ee943c8bcf36f1e2218d8e25edfa38ec5fe4bec2
829 Author: Samuel Just <sam.just@inktank.com>
830 Date: Wed Feb 27 16:58:45 2013 -0800
832 FileJournal::wrap_read_bl: adjust pos before returning
834 Otherwise, we may feed an offset past the end of the journal to
835 check_header in read_entry and incorrectly determine that the entry is
841 Reviewed-by: Sage Weil <sage@inktank.com>
842 Signed-off-by: Samuel Just <sam.just@inktank.com>
843 (cherry picked from commit 5d54ab154ca790688a6a1a2ad5f869c17a23980a)
845 commit 8c6f52215240f48b5e4d5bb99a5f2f451e7ce70a
846 Author: Sage Weil <sage@inktank.com>
847 Date: Wed Jan 16 13:14:00 2013 -0800
849 osd: leave osd_lock locked in shutdown()
851 No callers expect the lock to be dropped.
854 Signed-off-by: Sage Weil <sage@inktank.com>
855 (cherry picked from commit 98a763123240803741ac9f67846b8f405f1b005b)
857 commit a324d999804b4648f245ee36b3bd611b3d139d5d
858 Author: Sage Weil <sage@inktank.com>
859 Date: Tue Feb 26 14:07:12 2013 -0800
861 msg: fix entity_addr_t::is_same_host() for IPv6
863 We weren't checking the memcmp return value properly! Aie...
866 Signed-off-by: Sage Weil <sage@inktank.com>
867 (cherry picked from commit c8dd2b67b39a8c70e48441ecd1a5cc3c6200ae97)
869 commit 79d68ae8581141c3fb4cfafd76d5111ff009b762
870 Author: Sage Weil <sage@inktank.com>
871 Date: Sun Feb 17 22:35:50 2013 -0800
873 osd: requeue pg waiters at the front of the finished queue
875 We could have a sequence like:
881 in the finished queue. Op1 gets put on waiting_for_pg, the notify
882 creates the pg and requeues op1 (and the end), op2 is handled, and
883 finally op1 is handled. That breaks ordering; see #2947.
885 Instead, when we wake up a pg, queue the waiting messages at the front
886 of the dispatch queue.
888 Signed-off-by: Sage Weil <sage@inktank.com>
889 Reviewed-by: Samuel Just <sam.just@inktank.com>
890 (cherry picked from commit 56c5a07708d52de1699585c9560cff8b4e993d0a)
892 commit 79c4e7e91becc497843d96251776bdc176706aa0
893 Author: Sage Weil <sage@inktank.com>
894 Date: Sun Feb 17 20:49:52 2013 -0800
896 osd: pull requeued requests off one at a time
898 Pull items off the finished queue on at a time. In certain cases, an
899 event may result in new items betting added to the finished queue that
900 will be put at the *front* instead of the back. See latest incarnation
903 Note that this is a significant changed in behavior in that we can
904 theoretically starve if an event keeps resulting in new events getting
907 Signed-off-by: Sage Weil <sage@inktank.com>
908 Reviewed-by: Samuel Just <sam.just@inktank.com>
909 (cherry picked from commit f1841e4189fce70ef5722d508289e516faa9af6a)
911 commit 3a6ce5d0355beaa56199465e94666cae40bd8da1
912 Author: Sage Weil <sage@inktank.com>
913 Date: Thu Jan 17 22:00:42 2013 -0800
915 mds: open mydir after replay
917 In certain cases, we may replay the journal and not end up with the
918 dirfrag for mydir open. This is fine--we just need to open it up and
921 Signed-off-by: Sage Weil <sage@inktank.com>
922 (cherry picked from commit e51299fbce6bdc3d6ec736e949ba8643afc965ec)
924 commit 36ed407e0f939a9bca57c3ffc0ee5608d50ab7ed
925 Author: Greg Farnum <greg@inktank.com>
926 Date: Thu Feb 21 09:21:01 2013 -0800
928 mds: use inode_t::layout for dir layout policy
930 Remove the default_file_layout struct, which was just a ceph_file_layout,
931 and store it in the inode_t. Rip out all the annoying code that put this
934 To aid in this usage, add a clear_layout() function to inode_t.
936 Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
937 Signed-off-by: Greg Farnum <greg@inktank.com>
938 Reviewed-by: Sage Weil <sage@inktank.com>
940 commit b4fd196cae5ab76aeb8f82f0581d0a6a2133a3ed
941 Author: Sage Weil <sage@inktank.com>
942 Date: Sun Jan 20 21:53:37 2013 -0800
944 mds: parse ceph.*.layout vxattr key/value content
946 Use qi to parse a strictly formatted set of key/value pairs. Be picky
947 about whitespace. Any subset of recognized keys is allowed. Parse the
948 same set of keys as the ceph.*.layout.* vxattrs.
950 Signed-off-by: Sage Weil <sage@inktank.com>
951 (cherry picked from commit 5551aa5b3b5c2e9e7006476b9cd8cc181d2c9a04)
953 commit 5f92adca26b509aed5655e85ad53174fa7b574b9
954 Author: Yehuda Sadeh <yehuda@inktank.com>
955 Date: Mon Feb 18 09:10:43 2013 -0800
957 rgw: fix multipart uploads listing
961 Listing multipart uploads had a typo, and was requiring the
962 wrong resource (uploadId instead of uploads).
964 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
965 Reviewed-by: Greg Farnum <greg@inktank.com>
966 (cherry picked from commit db99fb4417b87301a69cb37b00c35c838b77197e)
968 commit a44df9343594099fecb3897df393249d3d1992e2
969 Author: Yehuda Sadeh <yehuda@inktank.com>
970 Date: Fri Feb 15 10:22:54 2013 -0800
972 rgw: don't copy object when it's copied into itself
977 When object copied into itself, object will not be fully copied: tail
978 reference count stays the same, head part is rewritten.
980 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
981 Reviewed-by: Greg Farnum <greg@inktank.com>
982 (cherry picked from commit 34f885be536d0ac89c10fd29b1518751d2ffc547)
984 commit 91d6be8353259ca30dc87062422e9ae334c3c344
985 Author: Samuel Just <sam.just@inktank.com>
986 Date: Tue Feb 19 10:49:33 2013 -0800
988 PG: remove weirdness log for last_complete < log.tail
990 In the case of a divergent object prior to log.tail,
991 last_complete may end up before log.tail.
995 Signed-off-by: Samuel Just <sam.just@inktank.com>
996 Reviewed-by: Sage Weil <sage@inktank.com>
997 (cherry picked from commit dbadb3e2921297882c5836c67ca32bb8ecdc75db)
1003 commit 7889c5412deaab1e7ba53d66814d0f25659229c1
1004 Author: James Page <james.page@ubuntu.com>
1005 Date: Mon Feb 18 16:24:54 2013 +0000
1007 Strip any trailing whitespace from rbd showmapped
1009 More recent versions of ceph append a bit of whitespace to the line
1010 after the name of the /dev/rbdX device; this causes the monitor check
1011 to fail as it can't find the device name due to the whitespace.
1013 This fix excludes any characters after the /dev/rbdN match.
1014 (cherry picked from commit ad84ea07cac5096de38b51b8fc452c99f016b8d8)
1016 commit 42fba772c3c0e2396124a9d6ec39297dd9af4cc9
1017 Merge: 290a352 ad6655e
1018 Author: Sage Weil <sage@newdream.net>
1019 Date: Sun Feb 17 21:52:55 2013 -0800
1021 Merge pull request #64 from dalgaaf/wip-bobtail-memleaks
1023 cherry-pick some memleak fixes from master to bobtail
1025 commit ad6655e976916d3b62e0c91fd469eeb49fe7da3e
1026 Author: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1027 Date: Fri Feb 8 17:17:59 2013 +0100
1029 rgw/rgw_rest.cc: fix 4K memory leak
1031 Fix 4K memory leak in case RGWClientIO::read() fails in
1032 read_all_chunked_input().
1034 Error from cppcheck was:
1035 Checking src/rgw/rgw_rest.cc...
1036 [src/rgw/rgw_rest.cc:688]: (error) Memory leak: data
1038 Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1039 (cherry picked from commit 89df090e04ef9fc5aae29122df106b0347786fab)
1041 commit 3de692753c28ddbeb3c86b51466be16e94c9c458
1042 Author: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1043 Date: Fri Feb 8 17:14:19 2013 +0100
1045 SyntheticClient.cc: fix some memory leaks in the error handling
1047 Fix some memory leaks in case of error handling due to failed
1048 client->open() calls.
1050 Error from cppcheck was:
1051 [src/client/SyntheticClient.cc:1980]: (error) Memory leak: buf
1052 [src/client/SyntheticClient.cc:2040]: (error) Memory leak: buf
1053 [src/client/SyntheticClient.cc:2090]: (error) Memory leak: buf
1054 (cherry picked from commit f0ba80756d1c3c313014ad7be18191981fb545be)
1056 commit f19d228c6a49222659c769099aaa4e755b80331d
1057 Author: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1058 Date: Fri Feb 8 16:57:20 2013 +0100
1060 rgw/rgw_xml.cc: fix realloc memory leak in error case
1062 Fix error from cppcheck:
1064 [src/rgw/rgw_xml.cc:212]: (error) Common realloc mistake: 'buf'
1065 nulled but not freed upon failure
1067 Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1068 (cherry picked from commit d48cc789ea075ba2745754035640ada4131b2119)
1070 commit b0c6be95b03d9f3dd2badcdcff359ae7bc9684f4
1071 Author: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1072 Date: Fri Feb 8 16:54:33 2013 +0100
1074 os/FileStore.cc: fix realloc memory leak in error case
1076 Fix error from cppcheck:
1078 [src/os/FileStore.cc:512]: (error) Common realloc mistake: 'fiemap'
1079 nulled but not freed upon failure
1081 Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1082 (cherry picked from commit c92a0f552587a232f66620170660d6b2ab6fb3a5)
1084 commit f3e5cedbc11005701ac0a8e70909a6372cd2fe6f
1085 Author: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1086 Date: Fri Feb 8 16:49:36 2013 +0100
1088 common/fiemap.cc: fix realloc memory leak
1090 Fix error from cppcheck:
1092 [src/common/fiemap.cc:73]: (error) Common realloc mistake: 'fiemap'
1093 nulled but not freed upon failure
1095 Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
1096 (cherry picked from commit f26f1470e7af36fa1eb8dc59c8a7c62c3c3a22ba)
1098 commit 290a352c3f9e241deac562e980ac8c6a74033ba6
1099 Author: Sage Weil <sage@inktank.com>
1100 Date: Thu Feb 14 11:37:57 2013 -0800
1102 osd/OSDCap: add unit test for parsing pools/objects with _ and -
1104 Hunting #4122, where a user saw
1106 2013-02-13 19:39:25.467916 7f766fdb4700 10 osd.0 10 session 0x2c8cc60 client.libvirt has caps osdcap[grant(object_prefix rbd^@children class-read),grant(pool libvirt^@pool^@test rwx)] 'allow class-read object_prefix rbd_children, allow pool libvirt-pool-test rwx'
1108 Signed-off-by: Sage Weil <sage@inktank.com>
1109 (cherry picked from commit 2ce28ef1d7f95e71e1043912dfa269ea3b0d1599)
1110 (cherry picked from commit a6534bc8a0247418d5263b765772d5266f99229c)
1112 commit f47916353547c77a65c7b70e609a9096f09c6d28
1113 Author: Sage Weil <sage@inktank.com>
1114 Date: Thu Feb 14 15:39:43 2013 -0800
1116 osd/OSDCap: tweak unquoted_word parsing in osd caps
1118 Newer versions of spirit (1.49.0-3.1ubuntu1.1 in quantal, in particular)
1119 dislike the construct with alnum and replace the - and _ with '\0' in the
1124 Signed-off-by: Sage Weil <sage@inktank.com>
1125 Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
1126 (cherry picked from commit 6c504d96c1e4fbb67578fba0666ca453b939c218)