1 commit ffc58b4ea6a3ea10e91050aaa28d14d87dea5656 (tag: refs/tags/v0.67.8, refs/remotes/gh/dumpling)
2 Author: Jenkins <jenkins@inktank.com>
3 Date: Thu May 1 11:18:24 2014 +0000
7 commit 4b16b70c53be83481efefcf394eca99c73bb9805
9 Author: Sage Weil <sage@inktank.com>
10 Date: Wed Apr 30 15:15:48 2014 -0700
12 Merge pull request #1743 from ceph/wip-mon-backports.dumpling
14 mon: OSDMonitor: HEALTH_WARN on 'mon osd down out interval == 0'
16 Reviewed-by: Sage Weil <sage@inktank.com>
18 commit fb0944e22acf6f8b6cefb59cc4c41dc48087bfd7
19 Author: Joao Eduardo Luis <joao.luis@inktank.com>
20 Date: Wed Apr 30 17:13:30 2014 +0100
22 mon: OSDMonitor: HEALTH_WARN on 'mon osd down out interval == 0'
24 A 'status' or 'health' request will return a HEALTH_WARN whenever the
25 monitor handling the request has the option set to zero.
29 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
30 (cherry picked from commit b2112d5087b449d3b019678cb266ff6fa897897e)
32 commit 5a6b35160417423db7c6ff892627f084ab610dfe
33 Author: Sandon Van Ness <sandon@inktank.com>
34 Date: Tue Mar 4 16:15:15 2014 -0800
36 Make symlink of librbd to qemu's folder so it can detect it.
40 Signed-off-by: Sandon Van Ness <sandon@inktank.com>
41 (cherry picked from commit 65f3354903fdbdb81468a84b8049ff19c00f91ba)
43 commit 735a90a95eea01dbcce5026758895117c2842627
44 Author: Yehuda Sadeh <yehuda@inktank.com>
45 Date: Fri Apr 25 14:11:27 2014 -0700
50 This fixes the radosgw side of issue #8202. Needed to cast value
51 to unsigned char, otherwise it'd get padded.
55 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
56 (cherry picked from commit bcf92c496aba0dfde432290fc2df5620a2767313)
58 commit 438b57890dfce04226d769389a601d35b74e11fe
59 Merge: c049967 476b929
60 Author: Sage Weil <sage@inktank.com>
61 Date: Fri Apr 25 16:00:24 2014 -0700
63 Merge pull request #1700 from xanpeng/patch-1
65 Fix error in mkcephfs.rst
67 Signed-off-by: Xan Peng <xanpeng@gmail.com>
68 Reviewed-by: Sage Weil <sage@inktank.com>
70 commit 476b929ecc5b7351a5be3024817b900976a90a3e
71 Author: xanpeng <xanpeng@gmail.com>
72 Date: Mon Apr 21 11:30:42 2014 +0800
76 There should be no blank between mount options.
78 commit c049967af829497f8a62e0cbbd6031f85ead8a59
79 Author: Josh Durgin <josh.durgin@inktank.com>
80 Date: Tue Apr 1 17:27:01 2014 -0700
82 auth: add rwlock to AuthClientHandler to prevent races
84 For cephx, build_authorizer reads a bunch of state (especially the
85 current session_key) which can be updated by the MonClient. With no
86 locks held, Pipe::connect() calls SimpleMessenger::get_authorizer()
87 which ends up calling RadosClient::get_authorizer() and then
88 AuthClientHandler::bulid_authorizer(). This unsafe usage can lead to
91 Program terminated with signal 11, Segmentation fault.
92 0x00007fa0d2ddb7cb in ceph::buffer::ptr::release (this=0x7f987a5e3070) at common/buffer.cc:370
93 370 common/buffer.cc: No such file or directory.
96 0x00007fa0d2ddb7cb in ceph::buffer::ptr::release (this=0x7f987a5e3070) at common/buffer.cc:370
97 0x00007fa0d2ddec00 in ~ptr (this=0x7f989c03b830) at ./include/buffer.h:171
98 ceph::buffer::list::rebuild (this=0x7f989c03b830) at common/buffer.cc:817
99 0x00007fa0d2ddecb9 in ceph::buffer::list::c_str (this=0x7f989c03b830) at common/buffer.cc:1045
100 0x00007fa0d2ea4dc2 in Pipe::connect (this=0x7fa0c4307340) at msg/Pipe.cc:907
101 0x00007fa0d2ea7d73 in Pipe::writer (this=0x7fa0c4307340) at msg/Pipe.cc:1518
102 0x00007fa0d2eb44dd in Pipe::Writer::entry (this=<value optimized out>) at msg/Pipe.h:59
103 0x00007fa0e0f5f9d1 in start_thread (arg=0x7f987a5e4700) at pthread_create.c:301
104 0x00007fa0de560b6d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:115
108 Error in `qemu-system-x86_64': invalid fastbin entry (free): 0x00007ff12887ff20
109 *** ======= Backtrace: =========
110 /lib/x86_64-linux-gnu/libc.so.6(+0x80a46)[0x7ff3dea1fa46]
111 /usr/lib/librados.so.2(+0x29eb03)[0x7ff3e3d43b03]
112 /usr/lib/librados.so.2(_ZNK9CryptoKey7encryptEP11CephContextRKN4ceph6buffer4listERS4_RSs+0x71)[0x7ff3e3d42661]
113 /usr/lib/librados.so.2(_Z21encode_encrypt_enc_blIN4ceph6buffer4listEEvP11CephContextRKT_RK9CryptoKeyRS2_RSs+0xfe)[0x7ff3e3d417de]
114 /usr/lib/librados.so.2(_Z14encode_encryptIN4ceph6buffer4listEEiP11CephContextRKT_RK9CryptoKeyRS2_RSs+0xa2)[0x7ff3e3d41912]
115 /usr/lib/librados.so.2(_ZN19CephxSessionHandler12sign_messageEP7Message+0x242)[0x7ff3e3d40de2]
116 /usr/lib/librados.so.2(_ZN4Pipe6writerEv+0x92b)[0x7ff3e3e61b2b]
117 /usr/lib/librados.so.2(_ZN4Pipe6Writer5entryEv+0xd)[0x7ff3e3e6c7fd]
118 /lib/x86_64-linux-gnu/libpthread.so.0(+0x7f8e)[0x7ff3ded6ff8e]
119 /lib/x86_64-linux-gnu/libc.so.6(clone+0x6d)[0x7ff3dea99a0d]
121 Fix this by adding an rwlock to AuthClientHandler. A simpler fix would
122 be to move RadosClient::get_authorizer() into the MonClient() under
123 the MonClient lock, but this would not catch all uses of other
124 Authorizer, e.g. for verify_authorizer() and it would serialize
125 independent connection attempts.
127 This mainly matters for cephx, but none and unknown can have the
128 global_id reset as well.
130 Partially-fixes: #6480
131 Backport: dumpling, emperor
132 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
133 (cherry picked from commit 2cc76bcd12d803160e98fa73810de2cb916ef1ff)
135 commit 2b4b00b76b245b1ac6f95e4537b1d1a4656715d5
136 Author: Josh Durgin <josh.durgin@inktank.com>
137 Date: Tue Apr 1 11:37:29 2014 -0700
139 pipe: only read AuthSessionHandler under pipe_lock
141 session_security, the AuthSessionHandler for a Pipe, is deleted and
142 recreated while the pipe_lock is held. read_message() is called
143 without pipe_lock held, and examines session_security. To make this
144 safe, make session_security a shared_ptr and take a reference to it
145 while the pipe_lock is still held, and use that shared_ptr in
148 This may have caused crashes like:
150 *** Error in `qemu-system-x86_64': invalid fastbin entry (free): 0x00007f42a4002de0 ***
151 ======= Backtrace: =========
152 /lib/x86_64-linux-gnu/libc.so.6(+0x80a46)[0x7f452f1f3a46]
153 /usr/lib/x86_64-linux-gnu/libnss3.so(PK11_FreeSymKey+0xa8)[0x7f452e72ff98]
154 /usr/lib/librados.so.2(+0x2a18cd)[0x7f453451a8cd]
155 /usr/lib/librados.so.2(_ZNK9CryptoKey7encryptEP11CephContextRKN4ceph6buffer4listERS4_RSs+0x71)[0x7f4534519421]
156 /usr/lib/librados.so.2(_Z21encode_encrypt_enc_blIN4ceph6buffer4listEEvP11CephContextRKT_RK9CryptoKeyRS2_RSs+0xfe)[0x7f453451859e]
157 /usr/lib/librados.so.2(_Z14encode_encryptIN4ceph6buffer4listEEiP11CephContextRKT_RK9CryptoKeyRS2_RSs+0xa2)[0x7f45345186d2]
158 /usr/lib/librados.so.2(_ZN19CephxSessionHandler23check_message_signatureEP7Message+0x246)[0x7f4534516866]
159 /usr/lib/librados.so.2(_ZN4Pipe12read_messageEPP7Message+0xdcc)[0x7f453462ecbc]
160 /usr/lib/librados.so.2(_ZN4Pipe6readerEv+0xa5c)[0x7f453464059c]
161 /usr/lib/librados.so.2(_ZN4Pipe6Reader5entryEv+0xd)[0x7f4534643ecd]
162 /lib/x86_64-linux-gnu/libpthread.so.0(+0x7f8e)[0x7f452f543f8e]
163 /lib/x86_64-linux-gnu/libc.so.6(clone+0x6d)[0x7f452f26da0d]
165 Partially-fixes: #6480
166 Backport: dumpling, emperor
167 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
168 (cherry picked from commit 1d74170a4c252f35968ccfbec8e432582e92f638)
170 commit 48895a46015c9d6d67543816f5a400c21aa206b1 (refs/remotes/gh/wip-objectcacher-flusher-dumpling)
171 Author: Sage Weil <sage@inktank.com>
172 Date: Fri Jan 3 12:51:15 2014 -0800
174 osdc/ObjectCacher: back off less during flush
176 In cce990efc8f2a58c8d0fa11c234ddf2242b1b856 we added a limit to avoid
177 holding the lock for too long. However, if we back off, we currently
178 wait for a full second, which is probably a bit much--we really just want
179 to give other threads a chance.
182 Signed-off-by: Sage Weil <sage@inktank.com>
183 (cherry picked from commit e2ee52879e9de260abbf5eacbdabbd71973a6a83)
185 commit f3b5ba6f25010291a2918bdd286f1b39570bb907
186 Author: Sage Weil <sage@inktank.com>
187 Date: Tue Oct 1 09:28:29 2013 -0700
189 osdc/ObjectCacher: limit writeback IOs generated while holding lock
191 While analyzing a log from Mike Dawson I saw a long stall while librbd's
192 objectcacher was starting lots (many hundreds) of IOs. Limit the amount of
193 time we spend doing this at a time to allow IO replies to be processed so
194 that the cache remains responsive.
196 I'm not sure this warrants a tunable (which we would need to add for both
197 libcephfs and librbd).
199 Signed-off-by: Sage Weil <sage@inktank.com>
200 (cherry picked from commit cce990efc8f2a58c8d0fa11c234ddf2242b1b856)
202 commit 06f27fc6446d47b853208357ec4277c5dc10d9fe
203 Author: Sage Weil <sage@inktank.com>
204 Date: Tue Apr 8 10:52:43 2014 -0700
206 os/FileStore: reset journal state on umount
208 We observed a sequence like:
211 - sets JournalingObjectStore applied_op_seq
214 - initiate commit with prevous applied_op_seq
217 - on replay commit, we fail assert op > committed_seq
219 Although strictly speaking the assert failure is harmless here, in general
220 we should not let state leak through from a previous mount into this
221 mount or else assertions are in general more difficult to reason about.
224 Signed-off-by: Sage Weil <sage@inktank.com>
225 (cherry picked from commit 4de49e8676748b6ab4716ff24fd0a465548594fc)
227 commit b29238729f87c73dfdcf16dddcf293577678dea2
228 Author: Yehuda Sadeh <yehuda@inktank.com>
229 Date: Tue Nov 5 14:54:20 2013 -0800
231 rgw: deny writes to a secondary zone by non-system users
234 We don't want to allow regular users to write to secondary zones,
235 otherwise we'd end up with data inconsistencies.
237 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
238 (cherry picked from commit 6961b5254f16ac3362c3a51f5490328d23640dbf)
243 commit 051a17eb008d75aa6b0737873318a2e7273501ab
244 Author: Sage Weil <sage@inktank.com>
245 Date: Sat Apr 5 16:58:55 2014 -0700
247 mon: wait for quorum for MMonGetVersion
249 We should not respond to checks for map versions when we are in the
250 probing or electing states or else clients will get incorrect results when
251 they ask what the latest map version is.
254 Signed-off-by: Sage Weil <sage@inktank.com>
255 (cherry picked from commit 67fd4218d306c0d2c8f0a855a2e5bf18fa1d659e)
257 commit 0716516da05eee967796fb71eb2f85c86afc40f1
258 Author: Yehuda Sadeh <yehuda@inktank.com>
259 Date: Wed Feb 19 08:59:07 2014 -0800
261 rgw: fix swift range response
265 The range response header was broken in swift.
267 Reported-by: Julien Calvet <julien.calvet@neurea.com>
268 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
269 (cherry picked from commit 0427f61544529ab4e0792b6afbb23379fe722de1)
271 commit 94a1deefcfe525a7e698a1ae70a3bb561b6157de
272 Author: Yehuda Sadeh <yehuda@inktank.com>
273 Date: Fri Nov 22 15:41:49 2013 -0800
275 rgw: don't log system requests in usage log
278 System requets should not be logged in the usage log.
280 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
281 (cherry picked from commit 42ef8ba543c7bf13c5aa3b6b4deaaf8a0f9c58b6)
283 commit 23fed8fc427e7077c61f86168a42f61a5f73867d
284 Author: Greg Farnum <greg@inktank.com>
285 Date: Fri Apr 4 16:06:05 2014 -0700
287 OSD: _share_map_outgoing whenever sending a message to a peer
289 This ensures that they get new maps before an op which requires them (that
290 they would then request from the monitor).
292 Signed-off-by: Greg Farnum <greg@inktank.com>
293 (cherry picked from commit 232ac1a52a322d163d8d8dbc4a7da4b6a9acb709)
295 commit c45e15fd5cbe57a34c743b2835ecc30ee5a43963
296 Author: Xihui He <xihuihe@gmail.com>
297 Date: Mon Dec 30 12:04:10 2013 +0800
299 msgr: fix rebind() race
300 stop the accepter and mark all pipes down before rebind to avoid race
304 Signed-off-by: Xihui He xihuihe@gmail.com
305 (cherry picked from commit f8e413f9c79a3a2a12801f5f64a2f612de3f06a0)
307 commit 3d31cf012a59e1fea8080b13bdc06c9021ba0656
308 Author: Samuel Just <sam.just@inktank.com>
309 Date: Tue Nov 26 13:20:21 2013 -0800
311 PG: retry GetLog() each time we get a notify in Incomplete
313 If for some reason there are no up OSDs in the history which
314 happen to have usable copies of the pg, it's possible that
315 there is a usable copy elsewhere on the cluster which will
316 become known to the primary if it waits.
319 Signed-off-by: Samuel Just <sam.just@inktank.com>
320 Reviewed-by: Greg Farnum <greg@inktank.com>
321 Reviewed-by: Sage Weil <sage@inktank.com>
322 (cherry picked from commit 964c8e978f86713e37a13b4884a6c0b9b41b5bae)
324 commit 1f80bbdf45439c7224ed52e4956973fc6d007848
325 Author: Sage Weil <sage@inktank.com>
326 Date: Mon Mar 17 15:37:44 2014 -0700
328 os/FileJournal: return errors on make_writeable() if reopen fails
330 This is why #7738 is resulting in a crash instead of an error.
332 Signed-off-by: Sage Weil <sage@inktank.com>
333 (cherry picked from commit aed074401d2834a5b04edd1b7f6b4f36336f6293)
335 commit 62d942294a54208cdc82aebf8b536d164cae5dc6
336 Author: Sage Weil <sage@inktank.com>
337 Date: Mon Mar 17 16:21:17 2014 -0700
339 mon/Paxos: commit only after entire quorum acks
341 If a subset of the quorum accepts the proposal and we commit, we will start
342 sharing the new state. However, the mon that didn't yet reply with the
343 accept may still be sharing the old and stale value.
345 The simplest way to prevent this is not to commit until the entire quorum
346 replies. In the general case, there are no failures and this is just fine.
347 In the failure case, we will call a new election and have a smaller quorum
348 of (live) nodes and will recommit the same value.
350 A more performant solution would be to have a separate message invalidate
351 the old state and commit once we have all invalidations and a majority of
352 accepts. This will lower latency a bit in the non-failure case, but not
353 change the failure case significantly. Later!
356 Signed-off-by: Sage Weil <sage@inktank.com>
357 Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
358 Reviewed-by: Greg Farnum <greg@inktank.com>
359 (cherry picked from commit fa1d957c115a440e162dba1b1002bc41fc1eac43)
361 commit 2160c72b393893896be581e89a42c4e37b79cb23
362 Author: Samuel Just <sam.just@inktank.com>
363 Date: Thu Mar 13 14:04:19 2014 -0700
365 PrioritizedQueue: cap costs at max_tokens_per_subqueue
367 Otherwise, you can get a recovery op in the queue which has a cost
368 higher than the max token value. It won't get serviced until all other
369 queues also do not have enough tokens and higher priority queues are
373 Signed-off-by: Samuel Just <sam.just@inktank.com>
374 (cherry picked from commit 2722a0a487e77ea2aa0d18caec0bdac50cb6a264)
376 commit 1dd42e0f75fe1f5363f92bd5a4506812e54b8fb5
377 Author: Dan Mick <dan.mick@inktank.com>
378 Date: Thu Apr 3 13:59:59 2014 -0700
380 Fix byte-order dependency in calculation of initial challenge
383 Signed-off-by: Dan Mick <dan.mick@inktank.com>
384 Reviewed-by: Sage Weil <sage@inktank.com>
385 (cherry picked from commit 4dc62669ecd679bc4d0ef2b996b2f0b45b8b4dc7)
387 commit c66b61f9dcad217429e4876d27881d9fb2e7666f
388 Author: Josh Durgin <josh.durgin@inktank.com>
389 Date: Tue Dec 31 17:00:06 2013 -0800
391 rbd: return 0 and an empty list when pool is entirely empty
393 rbd_list will return -ENOENT when no rbd_directory object
394 exists. Handle this in the cli tool and interpret it as success with
397 Add this to the release notes since it changes command line behavior.
400 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
401 (cherry picked from commit ac547a5b7dc94282f079aef78e66348d99d9d5e9)
407 commit 60b7aa96abfe09f7e9a263fa3f9b72c556dee8cb
408 Author: Josh Durgin <josh.durgin@inktank.com>
409 Date: Wed Nov 20 18:35:34 2013 -0800
411 test: use older names for module setup/teardown
413 setUp and tearDown require nosetests 0.11, but 0.10.4 is the latest on
414 centos. Rename to use the older aliases, which still work with newer
415 versions of nosetests as well.
418 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
419 Reviewed-by: Dan Mick <dan.mick@inktank.com>
420 (cherry picked from commit f753d56a9edba6ce441520ac9b52b93bd8f1b5b4)
422 commit b405bfa49ec31f0c6d8636c4bdde17ee1f81deb7
423 Author: Samuel Just <sam.just@inktank.com>
424 Date: Sun Nov 3 11:06:10 2013 -0800
426 OSD: don't clear peering_wait_for_split in advance_map()
428 I really don't know why I added this... Ops can be discarded from the
429 waiting_for_pg queue if we aren't primary simply because there must have
430 been an exchange of peering events before subops will be sent within a
431 particular epoch. Thus, any events in the waiting_for_pg queue must be
432 client ops which should only be seen by the primary. Peering events, on
433 the other hand, should only be discarded if we are in a new interval,
434 and that check might as well be performed in the peering wq.
437 Signed-off-by: Samuel Just <sam.just@inktank.com>
438 Reviewed-by: Greg Farnum <greg@inktank.com>
439 (cherry picked from commit 9ab513334c7ff9544bac07bd420c6d5d200cf535)
441 commit a498c940bd630cb103d17ad8532a11122439411d
442 Merge: 225fc97 80e0a0a
443 Author: Sage Weil <sage@inktank.com>
444 Date: Wed Apr 2 12:57:30 2014 -0700
446 Merge remote-tracking branch 'gh/wip-7888-dumpling' into dumpling
448 commit 225fc97f228490dfc13c2e4deed8fecffdb28c5e
449 Author: Samuel Just <sam.just@inktank.com>
450 Date: Tue Nov 5 21:48:53 2013 -0800
452 PG: fix operator<<,log_wierdness log bound warning
454 Split may cause holes such that head != tail and yet
458 Signed-off-by: Samuel Just <sam.just@inktank.com>
459 Reviewed-by: David Zafman <david.zafman@inktank.com>
460 (cherry picked from commit c6826c1e8a301b2306530c6e5d0f4a3160c4e691)
462 commit 26eeab43f3f703a25e7ba62c75d0382d15e38263
463 Author: Samuel Just <sam.just@inktank.com>
464 Date: Tue Nov 5 17:47:48 2013 -0800
466 PGLog::rewind_divergent_log: log may not contain newhead
468 Due to split, there may be a hole at newhead.
471 Signed-off-by: Samuel Just <sam.just@inktank.com>
472 Reviewed-by: David Zafman <david.zafman@inktank.com>
473 (cherry picked from commit f4648bc6fec89c870e0c47b38b2f13496742b10f)
475 commit 040abd75ad45bbcc05b24c9dddbd2026dd35e659
476 Author: Sage Weil <sage@inktank.com>
477 Date: Sat Mar 29 14:23:21 2014 -0700
479 qa/workunits/fs/misc/layout_vxattrs: ceph.file.layout is not listed
481 As of 08a3d6bd428c5e78dd4a10e6ee97540f66f9729c. A similar change was made
484 Signed-off-by: Sage Weil <sage@inktank.com>
485 (cherry picked from commit 4f9f7f878953b29cd5f56a8e0834832d6e3a9cec)
487 commit f5aa492a338ff711f3a45e4ebbf0d6b187b5f78e
488 Merge: fef70cb 84cb345
489 Author: Sage Weil <sage@inktank.com>
490 Date: Fri Mar 28 18:01:08 2014 -0700
492 Merge pull request #1519 from ceph/wip-6951-dumpling
494 rgw: reset objv tracker on bucket recreation
496 commit fef70cbb52cf1ad12db45998b38858d9bbc3360d
497 Merge: 9bfbce3 f443ff3
498 Author: Sage Weil <sage@inktank.com>
499 Date: Fri Mar 28 17:02:39 2014 -0700
501 Merge pull request #1559 from ceph/wip-7881-dumpling
505 Reviewed-by: Sage Weil <sage@inktank.com>
507 commit 80e0a0a8fee2f6f903f612734b2cc72eae703eae (refs/remotes/gh/wip-7888-dumpling)
508 Author: Sage Weil <sage@inktank.com>
509 Date: Thu Mar 27 21:33:21 2014 -0700
511 mon/MonClient: use keepalive2 to verify the mon session is live
513 Verify that the mon is responding by checking the keepalive2 reply
514 timestamp. We cannot rely solely on TCP timing out and returning an
518 Signed-off-by: Sage Weil <sage@inktank.com>
519 (cherry picked from commit 056151a6334c054505c54e59af40f203a0721f28)
521 commit 8723218379e80725a449b0594a4b15eb1c236b05
522 Author: Sage Weil <sage@inktank.com>
523 Date: Thu Mar 27 21:09:13 2014 -0700
525 msgr: add KEEPALIVE2 feature
527 This is similar to KEEPALIVE, except a timestamp is also exchanged. It is
528 sent with the KEEPALIVE, and then returned with the ACK. The last
529 received stamp is stored in the Connection so that it can be queried for
530 liveness. Since all of the users of keepalive are already regularly
531 triggering a keepalive, they can check the liveness at the same time.
535 Signed-off-by: Sage Weil <sage@inktank.com>
536 (cherry picked from commit d747d79fd5ea8662a809c5636dfd2eaaa9bf8f5d)
540 src/include/ceph_features.h
542 commit a2f0974f8b3567c0385494a0b2c828ade6ca8e1c
543 Author: Greg Farnum <greg@inktank.com>
544 Date: Wed Mar 26 15:58:10 2014 -0700
546 Pipe: rename keepalive->send_keepalive
548 Signed-off-by: Greg Farnum <greg@inktank.com>
549 (cherry picked from commit 38d4c71a456c1cc9a5044dbcae5378836a34484d)
551 commit 9bfbce30678742515025ca235c4443bb3a69199f
552 Author: Sage Weil <sage@inktank.com>
553 Date: Wed Mar 26 21:52:00 2014 -0700
555 client: pin Inode during readahead
557 Make sure the Inode does not go away while a readahead is in progress. In
562 - get actual read from cache, return
564 - call ObjectCacher::release_set() and get unclean > 0, assert
567 Backport: emperor, dumpling
568 Signed-off-by: Sage Weil <sage@inktank.com>
569 (cherry picked from commit f1c7b4ef0cd064a9cb86757f17118d17913850db)
571 commit 232445578a2c6d0fb974e55378057fce473095f7
572 Author: Sage Weil <sage@inktank.com>
573 Date: Fri Mar 28 12:34:07 2014 -0700
575 osdc/ObjectCacher: call read completion even when no target buffer
577 If we do no assemble a target bl, we still want to return a valid return
578 code with the number of bytes read-ahead so that the C_RetryRead completion
579 will see this as a finish and call the caller's provided Context.
581 Signed-off-by: Sage Weil <sage@inktank.com>
582 (cherry picked from commit 032d4ec53e125ad91ad27ce58da6f38dcf1da92e)
584 commit f443ff3006d41a7b0a2d7b649e8def0ffef6df12
585 Author: Samuel Just <sam.just@inktank.com>
586 Date: Wed Oct 30 16:54:39 2013 -0700
588 PGLog: remove obsolete assert in merge_log
590 This assert assumes that if olog.head != log.head, olog contains
591 a log entry at log.head, which may not be true since pg splitting
592 might have left the log with arbitrary holes.
594 Related: 0c2769d3321bff6e85ec57c85a08ee0b8e751bcb
595 Signed-off-by: Samuel Just <sam.just@inktank.com>
596 Reviewed-by: Sage Weil <sage@inktank.com>
597 (cherry picked from commit 353813b2e1a98901b876790c7c531f8a202c661d)
599 commit 636e53c0f4fc43e9bfc1c8e7214cab9e0b46a359
600 Author: Samuel Just <sam.just@inktank.com>
601 Date: Mon Sep 30 15:54:27 2013 -0700
603 PGLog: on split, leave log head alone
605 This way last_update doesn't go backwards.
608 Signed-off-by: Samuel Just <sam.just@inktank.com>
609 (cherry picked from commit 0c2769d3321bff6e85ec57c85a08ee0b8e751bcb)
611 commit a0a560a9f04311306a9784fa3c6ea2586d637f56
612 Merge: 466cd53 41d5e9a
613 Author: Sage Weil <sage@inktank.com>
614 Date: Wed Mar 26 17:18:24 2014 -0700
616 Merge pull request #1539 from ceph/wip-6910-dumpling
618 PG: don't query unfound on empty pgs
620 commit 41d5e9ab39e69c80bec1cb0627004c3fae6dc81d
621 Author: Samuel Just <sam.just@inktank.com>
622 Date: Tue Nov 26 19:17:59 2013 -0800
624 PG: don't query unfound on empty pgs
626 When the replica responds, it responds with a notify
627 rather than a log, which the primary then ignores since
628 it is already in the peer_info map. Rather than fix that
629 we'll simply not send queries to peers we already know to
630 have no unfound objects.
633 Signed-off-by: Samuel Just <sam.just@inktank.com>
634 Reviewed-by: Sage Weil <sage@inktank.com>
635 Reviewed-by: David Zafman <david.zafman@inktank.com>
636 (cherry picked from commit 838b6c8387087543ce50837277f7f6b52ae87d00)
638 commit 466cd536ed7e541a36d88bce43683a2d9e2ca283
639 Merge: 2ef0d6a c188949
640 Author: Sage Weil <sage@inktank.com>
641 Date: Fri Mar 21 14:53:23 2014 -0700
643 Merge pull request #1313 from ceph/dumpling-osd-subscribe
645 Dumpling backport: clean up osd subscriptions
647 commit 2ef0d6a25bf8c0cfb38768c157c29ba52295f3ca
648 Merge: 77e46d0 bdd96c6
649 Author: Sage Weil <sage@inktank.com>
650 Date: Fri Mar 21 14:52:20 2014 -0700
652 Merge pull request #1485 from ceph/wip-7212.dumpling
654 backport 7212 fixes to dumpling
656 commit 84cb345e4f12a9b1db5e384411492a9d88f17dd8
657 Author: Yehuda Sadeh <yehuda@inktank.com>
658 Date: Wed Feb 19 08:11:56 2014 -0800
660 rgw: reset objv tracker on bucket recreation
663 If we cannot create a new bucket (as it already existed), we need to
664 read the old bucket's info. However, this was failing as we were holding
665 the objv tracker that we created for the bucket creation. We need to
666 clear it, as subsequent read using it will fail.
668 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
669 (cherry picked from commit 859ed33ed7f9a96f4783dfb3e130d5eb60c622dd)
671 commit 77e46d0d7984f2d3ee0e15f27d2961a637c20b45
672 Author: Samuel Just <sam.just@inktank.com>
673 Date: Wed Nov 6 14:33:03 2013 -0800
675 ReplicatedPG: don't skip missing if sentries is empty on pgls
677 Formerly, if sentries is empty, we skip missing. In general,
678 we need to continue adding items from missing until we get
679 to next (returned from collection_list_partial) to avoid
683 Signed-off-by: Samuel Just <sam.just@inktank.com>
684 Reviewed-by: David Zafman <david.zafman@inktank.com>
685 (cherry picked from commit c7a30b881151e08b37339bb025789921e7115288)
687 commit bdd96c620f33fb8f48f30f8d543af3290e6c934a
688 Author: Sage Weil <sage@inktank.com>
689 Date: Sat Feb 15 08:59:51 2014 -0800
691 mon/Elector: bootstrap on timeout
693 Currently if an election times out we call a new
694 election. If we have never joined a quorum, bootstrap
695 instead. This is heavier weight, but captures the case
696 where, during bootstrap:
698 - a and b have learned each others' addresses
699 - everybody calls an election
700 - a and b form a quorum
701 - c loops trying to call an election, but is ignored
702 because a and b don't see its address in the monmap
705 ubuntu@teuthology:/var/lib/teuthworker/archive/sage-2014-02-14_13:50:04-ceph-deploy-wip-7212-sage-b-testing-basic-plana/83194
707 Signed-off-by: Sage Weil <sage@inktank.com>
708 (cherry picked from commit a4bcb1f8129a4ece97bd3419abf1ff45d260ad8e)
709 (cherry picked from commit 143ec0281aa8b640617a3fe19a430248ce3b514c)
711 commit 68fcc63c0423a2071b7b944ea6c3448282a78a09
712 Author: Sage Weil <sage@inktank.com>
713 Date: Fri Feb 14 11:25:52 2014 -0800
715 mon: tell MonmapMonitor first about winning an election
717 It is important in the bootstrap case that the very first paxos round
718 also codify the contents of the monmap itself in order to avoid any manner
719 of confusing scenarios where subsequent elections are called and people
720 try to recover and modify paxos without agreeing on who the quorum
723 Signed-off-by: Sage Weil <sage@inktank.com>
724 (cherry picked from commit ad7f5dd481a7f45dfe6b50d27ad45abc40950510)
725 (cherry picked from commit e073a062d56099b5fb4311be2a418f7570e1ffd9)
727 commit a3e57b7231cb28c2e0a896f747537ebdbe3a4e96
728 Author: Sage Weil <sage@inktank.com>
729 Date: Fri Feb 14 11:13:26 2014 -0800
731 mon: only learn peer addresses when monmap == 0
733 It is only safe to dynamically update the address for a peer mon in our
734 monmap if we are in the midst of the initial quorum formation (i.e.,
735 monmap.epoch == 0). If it is a later epoch, we have formed our initial
736 quorum and any and all monmap changes need to be agreed upon by the quorum
737 and committed via paxos.
740 Signed-off-by: Sage Weil <sage@inktank.com>
741 (cherry picked from commit 7bd2104acfeff0c9aa5e648d82ed372f901f767f)
742 (cherry picked from commit 1996fd89fb3165a63449b135e05841579695aabd)
744 commit 21ed54201bd4b0f02c07f6f96a63a5720057f011
745 Author: Joao Eduardo Luis <joao.luis@inktank.com>
746 Date: Mon Mar 17 14:37:09 2014 +0000
748 ceph.in: do not allow using 'tell' with interactive mode
750 This avoids a lot of hassle when dealing with to whom tell each command
751 on interactive mode, and even more so if multiple targets are specified.
753 As so, 'tell' commands should be used while on interactive mode instead.
755 Backport: dumpling,emperor
757 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
758 (cherry picked from commit e39c213c1d230271d23b74086664c2082caecdb9)
760 commit be0205c33ccbab3b6f105bdf4da114658a981557
761 Author: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
762 Date: Wed Mar 12 22:56:44 2014 +0100
764 RGWListBucketMultiparts: init max_uploads/default_max with 0
766 CID 717377 (#1 of 1): Uninitialized scalar field (UNINIT_CTOR)
767 2. uninit_member: Non-static class member "max_uploads" is not initialized
768 in this constructor nor in any functions that it calls.
769 4. uninit_member: Non-static class member "default_max" is not initialized
770 in this constructor nor in any functions that it calls.
772 Signed-off-by: Danny Al-Gaaf <danny.al-gaaf@bisect.de>
773 (cherry picked from commit b23a141d54ffb39958aba9da7f87544674fa0e50)
775 commit 6c3d4fbeb9cc50eca6eba12cfe7fed64b34eec3d
776 Author: Sage Weil <sage@inktank.com>
777 Date: Thu Mar 13 14:49:30 2014 -0700
779 ceph_test_rados: wait for commit, not ack
781 First, this is what we wanted in the first place
783 Second, if we wait for ACK, we may look at a user_version value that is
787 Signed-off-by: Sage Weil <sage@inktank.com>
788 (cherry picked from commit f2124c5846f1e9cb44e66eb2e957b8c7df3e19f4)
792 src/test/osd/RadosModel.h
794 commit 2daed5ff99dab238b696da5aba3816c4f5d763e8
795 Author: Josh Durgin <josh.durgin@inktank.com>
796 Date: Thu Mar 13 09:50:16 2014 -0700
798 test-upgrade-firefly: skip watch-notify system test
800 This also fails on mixed version clusters due to watch on a
801 non-existent object returning ENOENT in firefly and 0 in dumpling.
803 Reviewed-by: Sage Weil <sage.weil@inktank.com>
804 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
806 commit 90a21d8cf6df5fe14b2dc9b2c175983b6bb017ce
807 Author: Sage Weil <sage@inktank.com>
808 Date: Wed Mar 12 21:30:12 2014 -0700
810 qa/workunit/rados/test-upgrade-firefly: skip watch-notify test
812 A watch on a non-existent object now returns ENOENT in firefly; skip this
813 test as it will fail on a hybrid or upgraded cluster.
815 Signed-off-by: Sage Weil <sage@inktank.com>
817 commit 32fdca6d9c2e3b923db7f21568bd315ab2c1c4ad
818 Merge: 6700dd0 cd7986c
819 Author: Sage Weil <sage@inktank.com>
820 Date: Tue Mar 11 21:33:40 2014 -0700
822 Merge pull request #1411 from ceph/wip-7076-dumpling
824 dumpling backport of watchers check for rbd_remove()
826 commit 6700dd068e236473343d15eee6307d44156958a3
827 Author: Ray Lv <xiangyulv@gmail.com>
828 Date: Wed Feb 26 21:17:32 2014 +0800
830 rgw: off-by-one in rgw_trim_whitespace()
835 Reviewed-by: Yehuda Sadeh <yehuda@inktank.com>
836 Signed-off-by: Ray Lv <raylv@yahoo-inc.com>
837 (cherry picked from commit 195d53a7fc695ed954c85022fef6d2a18f68fe20)
839 commit cd7986caf6baee5f9d6498b113b3382e66dd6f77
840 Author: Ilya Dryomov <ilya.dryomov@inktank.com>
841 Date: Wed Jan 29 16:12:01 2014 +0200
843 rbd: check for watchers before trimming an image on 'rbd rm'
845 Check for watchers before trimming image data to try to avoid getting
846 into the following situation:
848 - user does 'rbd rm' on a mapped image with an fs mounted from it
849 - 'rbd rm' trims (removes) all image data, only header is left
850 - 'rbd rm' tries to remove a header and fails because krbd has a
851 watcher registered on the header
852 - at this point image cannot be unmapped because of the mounted fs
853 - fs cannot be unmounted because all its data and metadata is gone
855 Unfortunately, this fix doesn't make it impossible to happen (the
856 required atomicity isn't there), but it's a big improvement over the
859 Fixes: http://tracker.ceph.com/issues/7076
861 Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
862 Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
863 (cherry picked from commit 0a553cfa81b06e75585ab3c39927e307ec0f4cb6)
865 commit a931aaa6cc104d63b20c0cbe9e3af4006c3abfaf
866 Merge: b476d32 f5668b3
867 Author: Sage Weil <sage@inktank.com>
868 Date: Sun Mar 9 10:56:31 2014 -0700
870 Merge pull request #1407 from dachary/wip-7188-dumpling
872 common: ping existing admin socket before unlink (dumpling)
874 Reviewed-by: Sage Weil <sage@inktank.com>
876 commit f5668b363b0724f385bebded3cbc7f363893f985
877 Author: Loic Dachary <loic@dachary.org>
878 Date: Sat Feb 15 11:43:13 2014 +0100
880 common: ping existing admin socket before unlink
882 When a daemon initializes it tries to create an admin socket and unlinks
883 any pre-existing file, regardless. If such a file is in use, it causes
884 the existing daemon to loose its admin socket.
886 The AdminSocketClient::ping is implemented to probe an existing socket,
887 using the "0" message. The AdminSocket::bind_and_listen function is
888 modified to call ping() on when it finds existing file. It unlinks the
889 file only if the ping fails.
891 http://tracker.ceph.com/issues/7188 fixes: #7188
893 Backport: emperor, dumpling
894 Reviewed-by: Sage Weil <sage@inktank.com>
895 Signed-off-by: Loic Dachary <loic@dachary.org>
896 (cherry picked from commit 45600789f1ca399dddc5870254e5db883fb29b38)
898 commit b476d324c69d4e6018191a7ffea8c9d6c1dfa008
899 Merge: d3e13a7 5133dd6
900 Author: Sage Weil <sage@inktank.com>
901 Date: Wed Mar 5 14:19:31 2014 -0800
903 Merge pull request #1366 from ceph/wip-6820.dumpling
905 mon: OSDMonitor: don't crash if formatter is invalid during osd crush dump
907 commit d3e13a7cdab42fa33182680f45fe21b4f9dc4b20
908 Merge: c218c99 9c626e0
909 Author: Josh Durgin <josh.durgin@inktank.com>
910 Date: Wed Mar 5 12:45:57 2014 -0800
912 Merge pull request #1377 from ceph/wip-7584
914 qa/workunit/rados/test-upgrade-firely.sh
916 Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
918 commit 9c626e0b18f538eb60883da01713ba629220e35e
919 Author: Sage Weil <sage@inktank.com>
920 Date: Wed Mar 5 12:37:10 2014 -0800
922 qa/workunit/rados/test-upgrade-firely.sh
924 Skip the tests that don't pass when run against firefly OSDs.
927 Signed-off-by: Sage Weil <sage@inktank.com>
929 commit c218c999ecebe41e6de6fde76e85cc765cad8257
930 Merge: 0eabbf1 24711cd
931 Author: Samuel Just <sam.just@inktank.com>
932 Date: Tue Mar 4 07:28:44 2014 -0800
934 Merge pull request #1357 from ceph/wip-dumpling-removewq
936 OSD: ping tphandle during pg removal
938 Reviewed-by: Greg Farnum <greg@inktank.com>
940 commit 5133dd60e272d3fcbaacd5662a708ee4cf0db46d
941 Author: Joao Eduardo Luis <joao.luis@inktank.com>
942 Date: Fri Nov 22 02:17:16 2013 +0000
944 mon: OSDMonitor: don't crash if formatter is invalid during osd crush dump
946 Code would assume a formatter would always be defined. If a 'plain'
947 formatter or even an invalid formatter were to be supplied, the monitor
948 would crash and burn in poor style.
953 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
954 (cherry picked from commit 49d2fb71422fe4edfe5795c001104fb5bc8c98c3)
956 commit 24711cd49f85dbe827d41c4bcad2700cd6c42ad7
957 Author: Samuel Just <sam.just@inktank.com>
958 Date: Tue Oct 15 13:11:29 2013 -0700
960 OSD: ping tphandle during pg removal
963 Signed-off-by: Samuel Just <sam.just@inktank.com>
964 Reviewed-by: Sage Weil <sage@inktank.com>
966 (cherry picked from commit c658258d9e2f590054a30c0dee14a579a51bda8c)
971 commit 0eabbf145e1c44f4d128b192cc77b708f180c968
972 Merge: fe8915a 9d5d931
973 Author: Samuel Just <sam.just@inktank.com>
974 Date: Tue Feb 25 15:47:05 2014 -0800
976 Merge pull request #1316 from ceph/dumpling-6922
978 Dumpling: Prevent extreme PG split multipliers
980 Reviewed-by: Samuel Just <sam.just@inktank.com>
982 commit fe8915ae7e182340d1e22154e852895742c7da51
983 Merge: 87822cc 5667566
984 Author: Samuel Just <sam.just@inktank.com>
985 Date: Tue Feb 25 15:45:45 2014 -0800
987 Merge pull request #1315 from ceph/dumpling-hashpspool
989 mon: OSDMonitor: allow (un)setting 'hashpspool' flag via 'osd pool set'
991 Reviewed-by: Samuel Just <sam.just@inktank.com>
993 commit 87822ccc862b533132c1fe232dfe4b7b17b816ad
994 Merge: 0ae3352 37fbcb9
995 Author: Samuel Just <sam.just@inktank.com>
996 Date: Tue Feb 25 15:44:39 2014 -0800
998 Merge pull request #1314 from ceph/dumpling-osd-pgstatsack
1000 Dumpling osd pgstatsack
1002 Reviewed-by: Samuel Just <sam.just@inktank.com>
1004 commit 5667566313b69dca011e897b2fa752356ad8901b
1005 Author: Joao Eduardo Luis <joao.luis@inktank.com>
1006 Date: Thu Oct 10 17:43:48 2013 -0700
1008 mon: OSDMonitor: allow (un)setting 'hashpspool' flag via 'osd pool set'
1010 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
1011 Reviewed-by: Sage Weil <sage@inktank.com>
1012 (cherry picked from commit 1c2886964a0c005545abab0cf8feae7e06ac02a8)
1016 src/mon/MonCommands.h
1017 src/mon/OSDMonitor.cc
1019 mon: ceph hashpspool false clears the flag
1020 instead of toggling it.
1021 Signed-off-by: Loic Dachary <loic@dachary.org>
1022 Reviewed-by: Christophe Courtaut <christophe.courtaut@gmail.com>
1023 Reviewed-by: Sage Weil <sage@inktank.com>
1024 (cherry picked from commit 589e2fa485b94244c79079f249428d4d545fca18
1026 Replace some of the infrastructure required by this command that
1027 was not present in Dumpling with single-use code.
1028 Signed-off-by: Greg Farnum <greg@inktank.com>
1030 commit 9d5d931c60104823b3b20dcfb09480d65ffaa5ed
1031 Author: Greg Farnum <greg@inktank.com>
1032 Date: Tue Dec 3 10:57:09 2013 -0800
1034 OSDMonitor: use a different approach to prevent extreme multipliers on PG splits
1036 Signed-off-by: Greg Farnum <greg@inktank.com>
1037 Reviewed-by: Sage Weil <sage@inktank.com>
1038 (cherry picked from commit d8ccd73968fbd0753ca08916ebf1062cdb4d5ac1)
1042 src/mon/OSDMonitor.cc
1044 commit c0c4448dc7df7900a564a6745903398cd39be7f1
1045 Author: Greg Farnum <greg@inktank.com>
1046 Date: Mon Dec 2 15:13:40 2013 -0800
1048 OSDMonitor: prevent extreme multipliers on PG splits
1053 Signed-off-by: Greg Farnum <greg@inktank.com>
1054 Reviewed-by: Sage Weil <sage@inktank.com>
1055 (cherry picked from commit f57dad6461171c903e8b5255eaed300374b00e74)
1059 src/mon/OSDMonitor.cc
1061 commit c1889497b93ae9f0a946b11d9f5f6fcc7427e934
1062 Author: Sage Weil <sage@inktank.com>
1063 Date: Sat Feb 22 08:08:37 2014 -0800
1065 osd: fix off-by-one is boot subscription
1067 If we have osdmap N, we want to onetime subscribe
1068 starting at N+1. Among other things, it means we
1069 hear when the NOUP flag is cleared.
1071 This appears to have broken somewhere around
1072 3c76b81f2f96b790b72f2088164ed8e9d5efbba1.
1075 Signed-off-by: Sage Weil <sage@inktank.com>
1076 Reviewed-by: Sam Just <sam.just@inktank.com>
1077 (cherry picked from commit 70d23b9a0ad9af5ca35a627a7f93c7e610e17549)
1078 Reviewed-by: Greg Farnum <greg@inktank.com>
1080 commit 4584f60653bee0305e85418323d80332ceecd0cf
1081 Author: Greg Farnum <greg@inktank.com>
1082 Date: Tue Feb 11 12:51:19 2014 -0800
1084 OSD: use the osdmap_subscribe helper
1086 Signed-off-by: Greg Farnum <greg@inktank.com>
1087 Reviewed-by: Sage Weil <sage@inktank.com>
1088 (cherry picked from commit 3c76b81f2f96b790b72f2088164ed8e9d5efbba1)
1090 commit 61b2aeee7c37e03d5f6691c08c7760c48a85a2e1
1091 Author: Greg Farnum <greg@inktank.com>
1092 Date: Tue Feb 11 13:34:39 2014 -0800
1094 OSD: create a helper for handling OSDMap subscriptions, and clean them up
1096 We've had some trouble with not clearing out subscription requests and
1097 overloading the monitors (though only because of other bugs). Write a
1098 helper for handling subscription requests that we can use to centralize
1099 safety logic. Clear out the subscription whenever we get a map that covers
1100 it; if there are more maps available than we received, we will issue another
1101 subscription request based on "m->newest_map" at the end of handle_osd_map().
1103 Notice that the helper will no longer request old maps which we already have,
1104 and that unless forced it will not dispatch multiple subscribe requests
1105 to a single monitor.
1106 Skipping old maps is safe:
1107 1) we only trim old maps when the monitor tells us to,
1108 2) we do not send messages to our peers until we have updated our maps
1110 That means only old and broken OSDs will send us messages based on maps
1111 in our past, and we can (and should) ignore any directives from them anyway.
1113 Signed-off-by: Greg Farnum <greg@inktank.com>
1114 Reviewed-by: Sage Weil <sage@inktank.com>
1115 (cherry picked from commit 6db3ae851d1c936de045390d18b1c6ae95f2a209)
1121 commit d93d67d1a315d8abe8d1cd9d7ea83417a19e2406
1122 Author: Greg Farnum <greg@inktank.com>
1123 Date: Tue Feb 11 13:31:26 2014 -0800
1125 monc: new fsub_want_increment( function to make handling subscriptions easier
1127 Provide a subscription-modifying function which will not decrement
1130 Signed-off-by: Greg Farnum <greg@inktank.com>
1131 Reviewed-by: Sage Weil <sage@inktank.com>
1132 (cherry picked from commit 5b9c187caf6f7847aaa4a1003d200158dd32bf63)
1134 commit 37fbcb958f79bbfcba57c516b4862a14c52be398
1135 Author: Greg Farnum <greg@inktank.com>
1136 Date: Wed Feb 12 11:30:15 2014 -0800
1138 OSD: disable the PGStatsAck timeout when we are reconnecting to a monitor
1140 Previously, the timeout counter started as soon as we issued the reopen,
1141 but if the reconnect process itself took a while, we might time out and
1142 issue another reopen just as we get to the point where it's possible to
1143 get work done. Since the mon client has its own reconnect timeouts (that is,
1144 the OSD doesn't need to trigger those), we instead disable our timeouts
1145 while the reconnect is happening, and then turn them back on again starting
1146 from when we get the reconnect callback.
1148 Signed-off-by: Greg Farnum <greg@inktank.com>
1149 Reviewed-by: Sage Weil <sage@inktank.com>
1150 (cherry picked from commit 64cedf6fa3ee309cc96554286bfb805e4ca89439)
1156 commit 9be395e870c50d97604ef41f17667cc566fd84e1
1157 Author: Greg Farnum <greg@inktank.com>
1158 Date: Wed Feb 12 13:51:48 2014 -0800
1160 monc: backoff the timeout period when reconnecting
1162 If the monitors are systematically slowing down, we don't want to spam
1163 them with reconnect attempts every three seconds. Instead, every time
1164 we issue a reconnect, multiply our timeout period by a configurable; when
1165 we complete the connection, reduce that multipler by 50%. This should let
1166 us respond to monitor load.
1167 Of course, we don't want to do that for initial startup in the case of a
1168 couple down monitors, so don't apply the backoff until we've successfully
1169 connected to a monitor at least once.
1171 Signed-off-by: Greg Farnum <greg@inktank.com>
1172 Reviewed-by: Sage Weil <sage@inktank.com>
1173 (cherry picked from commit 794c86fd289bd62a35ed14368fa096c46736e9a2)
1175 commit 8f4c20bdab153d1603cc99186d8d3e3970aa8976
1176 Author: Greg Farnum <greg@inktank.com>
1177 Date: Wed Feb 12 13:37:50 2014 -0800
1179 monc: set "hunting" to true when we reopen the mon session
1181 If we don't have a connecton to a monitor, we want to retry to another
1182 monitor regardless of whether it's the first time or not.
1184 Signed-off-by: Greg Farnum <greg@inktank.com>
1185 Reviewed-by: Sage Weil <sage@inktank.com>
1186 (cherry picked from commit 60da8abe0ebf17ce818d6fcc6391401878123bb7)
1188 commit c6317558e0d3c8c62aecee0d95a839f93303f681
1189 Author: Greg Farnum <greg@inktank.com>
1190 Date: Tue Feb 11 17:53:56 2014 -0800
1192 monc: let users specify a callback when they reopen their monitor session
1194 Then the callback is triggered when a new session is established, and the
1195 daemon can do whatever it likes. There are no guarantees about how long it
1196 might take to trigger, though. In particular we call the provided callback
1197 while not holding our own lock in order to avoid deadlock. This could lead
1198 to some funny ordering from the user's perspective if they call
1199 reopen_session() again before getting the callback, but there's no way around
1200 that, so they just have to use it appropriately.
1202 Signed-off-by: Greg Farnum <greg@inktank.com>
1203 Reviewed-by: Sage Weil <sage@inktank.com>
1204 (cherry picked from commit 1a8c43474bf36bfcf2a94bf9b7e756a2a99f33fd)
1206 commit 0ae335298b85daba5125a3da4ad26d598c76ecab (refs/remotes/gh/multi-object-delete)
1207 Author: Yehuda Sadeh <yehuda@inktank.com>
1208 Date: Tue Feb 11 16:54:05 2014 -0800
1210 rgw: multi object delete should be idempotent
1213 When doing a multi object delete, if an object does not exist then we
1214 should return a success code for that object.
1216 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
1217 (cherry picked from commit 8ca3d95bf633ea9616852cec74f02285a03071d5)