1 commit 920f82e805efec2cae05b79c155c07df0f3ed5dd
2 Author: Gary Lowell <gary.lowell@inktank.com>
3 Date: Mon Jan 7 21:08:08 2013 -0800
7 commit 213e3559dd260a2e19324f2a671c808261249f96
8 Author: Sage Weil <sage@inktank.com>
9 Date: Sun Jan 6 20:43:21 2013 -0800
11 osd: fix race in do_recovery()
13 Verify that the PG is still RECOVERING or BACKFILL when we take the pg
14 lock in the recovery thread. This prevents a crash from an invalid
15 state machine event when the recovery queue races with a PG state change
16 (e.g., due to peering).
18 Signed-off-by: Sage Weil <sage@inktank.com>
19 Reviewed-by: Samuel Just <sam.just@inktank.com>
21 commit e410d1a066b906cad3103a5bbfa5b4509be9ac37
22 Author: Samuel Just <sam.just@inktank.com>
23 Date: Fri Jan 4 21:19:45 2013 -0800
25 ReplicatedPG: requeue waiting_for_ondisk in apply_and_flush_repops
28 Signed-off-by: Samuel Just <sam.just@inktank.com>
29 Reviewed-by: Sage Weil <sage@inktank.com>
31 commit cd5f2bfdbb7fbf9237bef482845644cc41fa66de
32 Author: Josh Durgin <josh.durgin@inktank.com>
33 Date: Thu Nov 15 16:20:33 2012 -0800
35 ObjectCacher: fix off-by-one error in split
37 This error left a completion that should have been attached
38 to the right BufferHead on the left BufferHead, which would
39 result in the completion never being called unless the buffers
40 were merged before it's original read completed. This would cause
41 a hang in any higher level waiting for a read to complete.
43 The existing loop went backwards (using a forward iterator),
44 but stopped when the iterator reached the beginning of the map,
45 or when a waiter belonged to the left BufferHead.
47 If the first list of waiters should have been moved to the right
48 BufferHead, it was skipped because at that point the iterator
49 was at the beginning of the map, which was the main condition
52 Restructure the waiters-moving loop to go forward in the map instead,
53 so it's harder to make an off-by-one error.
56 Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
57 (cherry picked from commit 2e862f4d183d8b57b43b0777737886f18f68bf00)
59 commit 802c486fd0913e9fe58ae7ff72df135f99712358
60 Author: Sage Weil <sage@inktank.com>
61 Date: Fri Jan 4 11:07:48 2013 -0800
63 config: change default log_max_recent to 10,000
65 Commit c34e38bcdc0460219d19b21ca7a0554adf7f7f84 meant to do this but got
66 the wrong number of zeros.
68 Signed-off-by: Sage Weil <sage@inktank.com>
70 commit 39a734fbf34ccd121f17023bcec814e61c8bdaab
71 Author: Sage Weil <sage@inktank.com>
72 Date: Thu Jan 3 17:15:07 2013 -0800
74 os/FileStore: fix non-btrfs op_seq commit order
76 The op_seq file is the starting point for journal replay. For stable btrfs
77 commit mode, which is using a snapshot as a reference, we should write this
78 file before we take the snap. We normally ignore current/ contents anyway.
80 On non-btrfs file systems, however, we should only write this file *after*
81 we do a full sync, and we should then fsync(2) it before we continue
82 (and potentially trim anything from the journal).
84 This fixes a serious bug that could cause data loss and corruption after
85 a power loss event. For a 'kill -9' or crash, however, there was little
86 risk, since the writes were still captured by the host's cache.
89 Signed-off-by: Sage Weil <sage@inktank.com>
90 Reviewed-by: Samuel Just <sam.just@inktank.com>
91 (cherry picked from commit 28d59d374b28629a230d36b93e60a8474c902aa5)
93 commit 49416619d733572368e5d2ba7f2b34150c754b23
94 Author: Sage Weil <sage@inktank.com>
95 Date: Fri Dec 28 13:07:18 2012 -0800
97 log: broadcast cond signals
99 We were using a single cond, and only signalling one waiter. That means
100 that if the flusher and several logging threads are waiting, and we hit
101 a limit, we the logger could signal another logger instead of the flusher,
102 and we could deadlock.
104 Similarly, if the flusher empties the queue, it might signal only a single
105 logger, and that logger could re-signal the flusher, and the other logger
108 Intead, break the single cond into two: one for loggers, and one for the
109 flusher. Always signal the (one) flusher, and always broadcast to all
112 Backport: bobtail, argonaut
113 Signed-off-by: Sage Weil <sage@inktank.com>
114 Reviewed-by: Dan Mick <dan.mick@inktank.com>
115 (cherry picked from commit 813787af3dbb99e42f481af670c4bb0e254e4432)
117 commit f8bb4814d0fd7e08c9b6cdcde02940bf1b42cadd
118 Author: Sage Weil <sage@inktank.com>
119 Date: Wed Jan 2 13:58:44 2013 -0800
121 log: fix locking typo/stupid for dump_recent()
123 We weren't locking m_flush_mutex properly, which in turn was leading to
124 racing threads calling dump_recent() and garbling the crash dump output.
126 Backport: bobtail, argonaut
127 Signed-off-by: Sage Weil <sage@inktank.com>
128 Reviewed-by: Dan Mick <dan.mick@inktank.com>
129 (cherry picked from commit 43cba617aa0247d714632bddf31b9271ef3a1b50)
131 commit 9483a032f750572586f146c696ec6501d3df0383
132 Author: Sage Weil <sage@inktank.com>
133 Date: Thu Dec 27 16:06:24 2012 -0800
135 init-ceph: fix status version check across machines
137 The local state isn't propagated into the backtick shell, resulting in
138 'unknown' for all remote daemons. Avoid backticks altogether.
140 Signed-off-by: Sage Weil <sage@inktank.com>
141 (cherry picked from commit 6c7b667badc5e7608b69c533a119a2afc062e257)
143 commit 8fef9360fe949fbf4ee6e468c36cc1c84b369b6e
144 Author: Travis Rhoden <trhoden@gmail.com>
145 Date: Mon Aug 20 13:29:11 2012 -0700
147 init-ceph: use SSH in "service ceph status -a" to get version
149 When running "service ceph status -a", a version number was never
150 returned for remote hosts, only for the local. This was because
151 the command to query the version number didn't use the do_cmd
152 function, which is responsible for running the command over SSH
155 Modify the ceph init.d script to use do_cmd for querying the
158 Signed-off-by: Travis Rhoden <trhoden@gmail.com>
159 (cherry picked from commit 60fdb6fda6233b01dae4ed8a34427d5960840b84)
161 commit c34e38bcdc0460219d19b21ca7a0554adf7f7f84
162 Author: Sage Weil <sage@inktank.com>
163 Date: Wed Nov 28 13:00:36 2012 -0800
165 log: 10,000 recent log entries
167 This is what we were (wrongly) doing before, so there are no memory
168 utilization surprises.
170 Signed-off-by: Sage Weil <sage@inktank.com>
171 (cherry picked from commit 78286b1403a5e0f14f95fe6b92f2fdb163e909f1)
173 commit 4daede79f578cccd340210752f912e685b1ff03d
174 Author: Sage Weil <sage@inktank.com>
175 Date: Wed Nov 28 12:59:43 2012 -0800
177 log: fix log_max_recent config
181 Signed-off-by: Sage Weil <sage@inktank.com>
182 (cherry picked from commit 4de7748b72d4f90eb1197a70015c199c15203354)
184 commit fdae0552a0b8b1a634775609b473aff9221904f3
185 Author: Sage Weil <sage@inktank.com>
186 Date: Thu Dec 20 13:48:06 2012 -0800
188 log: fix flush/signal race
190 We need to signal the cond in the same interval where we hold the lock
191 *and* modify the queue. Otherwise, we can have a race like:
193 queue has 1 item, max is 1.
194 A: enter submit_entry, signal cond, wait on condition
195 B: enter submit_entry, signal cond, wait on condition
196 C: flush wakes up, flushes 1 previous item
197 A: retakes lock, enqueues something, exits
198 B: retakes lock, condition fails, waits
199 -> C is never woken up as there are 2 items waiting
201 Signed-off-by: Sage Weil <sage@inktank.com>
202 Reviewed-by: Dan Mick <dan.mick@inktank.com>
203 (cherry picked from commit 50914e7a429acddb981bc3344f51a793280704e6)
205 commit bdcf6647dec05617d3da46ad00459498ede94f1d
206 Author: Gary Lowell <gary.lowell@inktank.com>
207 Date: Fri Dec 21 17:12:07 2012 -0800
209 .gitignore: Add ar-lib to ignore list
211 commit 51a900cfd7cc009f03fdb80be028015fc34547d2
212 Author: Gary Lowell <gary.lowell@inktank.com>
213 Date: Fri Dec 21 16:55:27 2012 -0800
215 autogen.sh: Create m4 directory for leveldb
217 Signed-off-by: Gary Lowell <gary.lowell@inktank.com>
219 commit 74473bb667df0ee0bb1adc6c43242733b80aa10d
220 Author: Gary Lowell <gary.lowell@inktank.com>
221 Date: Fri Dec 21 16:17:33 2012 -0800
223 leveldb: Update submodule
225 Signed-off-by: Gary Lowell <gary.lowell@inktank.com>
227 commit bc4f74c7b9ec4e6bbcf4ab4ab68a82a69b8eaca5
228 Author: Gary Lowell <gary.lowell@inktank.com>
229 Date: Thu Dec 20 16:49:32 2012 -0800
231 ceph.spec.in: Fedora builds debuginfo by default.
233 Signed-off-by: Gary Lowell <gary.lowell@inktank.com>
235 commit 3ed2d59e98494a0da29f4bb8537471237a6f8e93
236 Author: Yehuda Sadeh <yehuda@inktank.com>
237 Date: Wed Dec 19 17:07:18 2012 -0800
239 rgw: fix error handling with swift
242 verify_swift_token returns a bool and not an int.
244 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
246 commit dfd310368267df848f2b65cc536b4ffcb039d353
247 Author: Sam Lang <sam.lang@inktank.com>
248 Date: Mon Sep 24 09:55:25 2012 -0700
250 client: Fix for #3184 cfuse segv with no keyring
252 Fixes bug #3184 where the ceph-fuse client segfaults if authx is
253 enabled but no keyring file is present. This was due to the
254 client->init() return value not getting checked.
256 Signed-off-by: Sam Lang <sam.lang@inktank.com>
257 (cherry picked from commit 47983df4cbd31f299eef896b4612d3837bd7c7bd)
259 commit acebcce91448421c34a72850a380fcd1aabe9f41
260 Author: Joao Eduardo Luis <joao.luis@inktank.com>
261 Date: Tue Oct 9 21:25:54 2012 +0100
263 mon: Monitor: resolve keyring option to a file before loading keyring
265 Otherwise our keyring default location, or any other similarly formatted
266 location, will be taken as the actual location for the keyring and fail.
268 Reported-by: tziOm (at) #ceph
272 Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
273 (cherry picked from commit 7ef0df25e001bfae303feb3ae36514608767b1f2)
275 commit 7ab00a796d2233c3804f52fbba16c2f3ed0c297b
276 Author: Gary Lowell <glowell@inktank.com>
277 Date: Wed Dec 5 19:39:11 2012 -0800
279 .gitignore: Add m4 macro directories to ignore list
281 commit b781ac00e16fd802b6e055dd9d453271db6efa39
282 Author: Gary Lowell <glowell@inktank.com>
283 Date: Thu Nov 8 12:43:24 2012 -0800
285 build: Add RPM release string generated from git describe.
287 Fix for bug 3451. Use the commit count and sha1 from git describe to
288 construct a release string for rpm packages.
294 commit f60c9b6600a85b840a02e145bac77326d60fd03f
295 Author: Gary Lowell <glowell@inktank.com>
296 Date: Fri Nov 9 13:28:13 2012 -0800
298 ceph.spec.in: Build debuginfo subpackage.
300 This is a partial fix for bug 3471. Enable building of debuginfo package.
301 Some distributions enable this automatically by installing additional rpm
302 macros, on others it needs to be explicity added to the spec file.
304 commit 7fbf9476cb44ce256f4bb2366f46fa20cb570eda
305 Author: Yehuda Sadeh <yehuda@inktank.com>
306 Date: Mon Dec 3 14:32:28 2012 -0800
308 rgw: fix swift auth concurrency issue
311 Originally ops were using static structures, but that
312 has since changed. Switching swift auth handler to do
315 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
317 commit 0639cd9c479d69b077175f0385eb569ebb839349
318 Author: Yehuda Sadeh <yehuda@inktank.com>
319 Date: Thu Nov 29 13:39:22 2012 -0800
321 rgw: fix rgw_tools get_obj()
323 The original implementation broke whenever data exceeded
324 the chunk size. Also don't keep cache for objects that
325 exceed the chunk size as cache is not designed for
326 it. Increased chunk size to 512k.
328 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
330 commit 84983387a220a29540aca64af774bbe7bc6b5ee6
331 Author: Yehuda Sadeh <yehuda@inktank.com>
332 Date: Thu Nov 29 12:47:59 2012 -0800
336 This fixes a regression introduced at
337 17e4c0df44781f5ff1d74f3800722452b6a0fc58. The original
338 patch fixed error leak, however it also removed the
339 operation's send_response() call.
341 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
343 commit 6c1e2b663227cadff4255db3321976882df6b261
344 Author: Yehuda Sadeh <yehuda@inktank.com>
345 Date: Mon Nov 19 17:10:11 2012 -0800
347 rgw: fix xml parser leak
349 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
350 (cherry picked from commit f86522cdfcd81b2d28c581ac8b8de6226bc8d1a4)
352 commit 17e4c0df44781f5ff1d74f3800722452b6a0fc58
353 Author: Yehuda Sadeh <yehuda@inktank.com>
354 Date: Mon Nov 19 16:52:38 2012 -0800
356 rgw: fix memory leaks
358 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
359 (cherry picked from commit 98a04d76ebffa61c3ba4b033cdd57ac57b2f29f3)
365 commit cd18b9f43cce68ac3d8f109e82a5b603f9caad9b
366 Author: Yehuda Sadeh <yehuda@inktank.com>
367 Date: Wed Nov 7 13:21:15 2012 -0800
369 rgw: don't convert object mtime to UTC
372 When we read object info, don't try to convert mtime to
373 UTC, it's already in UTC.
375 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
377 commit ddb40dec6e8d930fd63a71ef1c98279d6472e3e4
378 Author: Yehuda Sadeh <yehuda@inktank.com>
379 Date: Wed Nov 14 11:30:34 2012 -0800
381 rgw: relax date format check
383 Don't try to parse beyond the GMT or UTC. Some clients use
384 special date formatting. If we end up misparsing the date
385 it'll fail in the authorization, so don't need to be too
388 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
390 commit e35e7e2bcb749e3ee00f1f93bc6b17bde8329414
391 Author: Sage Weil <sage@inktank.com>
392 Date: Tue Oct 30 14:17:56 2012 -0700
394 ceph-disk-activate: avoid duplicating mounts if already activated
396 If the given device is already mounted at the target location, do not
397 mount --move it again and create a bunch of dup entries in the /etc/mtab
398 and kernel mount table.
400 Signed-off-by: Sage Weil <sage@inktank.com>
401 (cherry picked from commit c435d314caeb5424c1f4482ad02f8a085317ad5b)
403 commit 0c9bbb3b4b4dbe6f0a12960785e35af9c748701a
404 Author: Sage Weil <sage@inktank.com>
405 Date: Thu Oct 25 21:21:18 2012 -0700
407 ceph-disk-prepare: poke kernel into refreshing partition tables
409 Prod the kernel to refresh the partition table after we create one. The
410 partprobe program is packaged with parted, which we already use, so this
411 introduces no new dependency.
413 Signed-off-by: Sage Weil <sage@inktank.com>
414 (cherry picked from commit 402e1f5319a52c309eca936081fddede1f107268)
416 commit b4004806ec387915ba4e039b9bc1c563273ef08f
417 Author: Sage Weil <sage@inktank.com>
418 Date: Thu Oct 25 21:20:21 2012 -0700
420 ceph-disk-prepare: fix journal partition creation
422 The end value needs to have + to indicate it is relative to wherever the
425 Signed-off-by: Sage Weil <sage@inktank.com>
426 (cherry picked from commit 2e32a0ee2d9e2a3bf5b138f50efc5fba8d5b8660)
428 commit efac1bb19f1ca3851c7b51d96cb4e82d8d142a45
429 Author: Sage Weil <sage@inktank.com>
430 Date: Thu Oct 25 18:14:47 2012 -0700
432 ceph-disk-prepare: assume parted failure means no partition table
434 If the disk has no valid label we get an error like
436 Error: /dev/sdi: unrecognised disk label
438 Assume any error we get is that and go with an id label of 1.
440 Signed-off-by: Sage Weil <sage@inktank.com>
441 (cherry picked from commit 8921fc7c7bc28fb98334c06f1f0c10af58085085)
443 commit 39b9da403541617fb088aa4ae03925850aa2de65
444 Merge: 83ab366 8c115d3
445 Author: Sage Weil <sage@inktank.com>
446 Date: Mon Nov 12 11:24:00 2012 -0800
448 Merge remote-tracking branch 'gh/wip-mds-stable' into stable
450 commit 8c115d3e81543a4c723bae217ba5b9d2c0e988f7
451 Author: Sage Weil <sage@inktank.com>
452 Date: Fri Nov 9 05:28:12 2012 -0800
454 mds: re-try_set_loner() after doing evals in eval(CInode*, int mask)
456 Consider a case where current loner is A and wanted loner is B.
457 At the top of the function we try to set the loner, but that may fail
458 because we haven't processed the gathered caps yet for the previous
459 loner. In the body we do that and potentially drop the old loner, but we
460 do not try_set_loner() again on the desired loner.
462 Try after our drop. If it succeeds, loop through the eval's one more time
463 so that we can issue caps approriately.
465 This fixes a hang induced by a simple loop like:
467 while true ; do echo asdf >> mnt.a/foo ; tail mnt.b/foo ; done &
468 while true ; do ls mnt.a mnt.b ; done
470 (The second loop may not be necessary.)
472 Signed-off-by: Sage Weil <sage@inktank.com>
474 commit 1a23bd6981e2a308be4a905021453a6d5feac9e3
475 Author: Samuel Just <sam.just@inktank.com>
476 Date: Fri Jul 13 14:23:27 2012 -0700
478 CompatSet: users pass bit indices rather than masks
480 CompatSet users number the Feature objects rather than
481 providing masks. Thus, we should do
483 mask |= (1 << f.id) rather than mask |= f.id.
485 In order to detect old, broken encodings, the lowest
486 bit will be set in memory but not set in the encoding.
487 We can reconstruct the correct mask from the names map.
489 This bug can cause an incompat bit to not be detected
494 Signed-off-by: Samuel Just <sam.just@inktank.com>
496 commit 83ab36655f397b7d781a611b4d82679c8484710f
497 Author: Gary Lowell <glowell@inktank.com>
498 Date: Wed Nov 7 12:41:10 2012 -0800
500 ceph.spec.in: Remove ceph version requirement from ceph-fuse package.
502 The ceph-fuse rpm package now only requires ceph as a pre-req, not a specific
505 commit 578b5764d5b26d190c295da88c1be5f0c77682a7
506 Author: Yehuda Sadeh <yehuda@inktank.com>
507 Date: Wed Oct 24 13:15:46 2012 -0700
509 rgw: fix multipart overwrite
512 Removed a few lines of code that prematurely created the head
513 part of the final object (before creating the manifest).
516 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
518 commit 96888ddf154231daa6c7138bb2b97674771cc610
519 Author: Sage Weil <sage@inktank.com>
520 Date: Mon Nov 5 23:27:13 2012 -0800
522 mds: move to from loner -> mix if *anyone* wants rd|wr
524 We were either going to MIX or SYNC depending on whether non-loners wanted
525 to read/write, but it may be that the loner wants to if our logic for
526 choosing loner vs not longer is based on anything other that just rd|wr
529 Signed-off-by: Sage Weil <sage@inktank.com>
531 commit b940d7750d14285fc85ce63b0f0444b58ad3ad93
532 Author: Sage Weil <sage@inktank.com>
533 Date: Mon Nov 5 23:26:09 2012 -0800
535 mds: base loner decision on wanted RD|WR|EXCL, not CACHE|BUFFER
537 Observed instance where one client wanted the Fc cap and prevented the
538 loner from getting RD|WR caps.
540 Signed-off-by: Sage Weil <sage@inktank.com>
542 commit 7a82a3214a38cbc7b5750adba23b760a4f51643a
543 Author: Sage Weil <sage@inktank.com>
544 Date: Tue Oct 30 09:00:11 2012 -0700
546 osd: make pool_snap_info_t encoding backward compatible
548 Way back in fc869dee1e8a1c90c93cb7e678563772fb1c51fb (v0.42) when we redid
549 the osd type encoding we forgot to make this conditionally encode the old
550 format for old clients. In particular, this means that kernel clients
551 will fail to decode the osdmap if there is a rados pool with a pool-level
555 Signed-off-by: Sage Weil <sage@inktank.com>
559 src/test/encoding/types.h
561 commit 8d7ca76182fdc04aaf8b698afbbd2d9b22e07fdc
562 Author: Yan, Zheng <zheng.z.yan@intel.com>
563 Date: Fri Sep 7 13:49:27 2012 +0800
565 osd/OSD.cc: Fix typo in OSD::heartbeat_check()
567 The check 'p->second.last_tx > cutoff' should always be false
568 since last_tx is periodically updated by OSD::heartbeat()
570 Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
571 Signed-off-by: Sage Weil <sage@inktank.com>
573 commit fc2b3ced4f7ed8d6e065c159fbeb8670fcdaca2a
574 Author: Yehuda Sadeh <yehuda@inktank.com>
575 Date: Mon Oct 22 16:52:11 2012 -0700
577 rgw: dump an error message if FCGX_Accept fails
579 Adding missing debug info.
581 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
583 commit d9b51b3b7011f4551fe27e33301b26cd23a72378
584 Author: Sage Weil <sage@inktank.com>
585 Date: Mon Oct 22 15:38:30 2012 -0700
587 workqueue: make debug output include active threads
589 Include active thread count in threadpool debug output.
591 Signed-off-by: Sage Weil <sage@inktank.com>
593 commit 852e27d3e19239a696fa693e2a30c3226d27a611
594 Author: Yehuda Sadeh <yehuda@inktank.com>
595 Date: Mon Oct 22 13:16:59 2012 -0700
597 rgw: don't continue processing of GET request on error
600 We continued processing requests long after the client
601 has died. This fix appliese to both s3 and swift.
604 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
606 commit 5f2b9b451da5a96fdbf9cf292648feb40dc13d28
607 Author: Sage Weil <sage@inktank.com>
608 Date: Fri Oct 19 08:46:19 2012 -0700
610 osd: be quiet about watches
614 Signed-off-by: Sage Weil <sage@inktank.com>
616 commit ea42eb1f7866fed0e843257ce05e34503fd0129c
617 Author: Sage Weil <sage@inktank.com>
618 Date: Wed Oct 17 17:44:12 2012 -0700
620 addr_parsing: make , and ; and ' ' all delimiters
622 Instead of just ,. Currently "foo.com, bar.com" will fail because of the
623 space after the comma. This patches fixes that, and makes all delim
624 chars interchangeable.
626 Signed-off-by: Sage Weil <sage@inktank.com>
628 commit b0164d9902c834c95187f6b277f58a30d26cc4d3
629 Author: Tommi Virtanen <tv@inktank.com>
630 Date: Fri Oct 5 10:57:42 2012 -0700
632 ceph-disk-prepare, debian/control: Support external journals.
634 Previously, ceph-disk-* would only let you use a journal that was a
635 file inside the OSD data directory. With this, you can do:
637 ceph-disk-prepare /dev/sdb /dev/sdb
639 to put the journal as a second partition on the same disk as the OSD
640 data (might save some file system overhead), or, more interestingly:
642 ceph-disk-prepare /dev/sdb /dev/sdc
644 which makes it create a new partition on /dev/sdc to use as the
645 journal. Size of the partition is decided by $osd_journal_size.
646 /dev/sdc must be a GPT-format disk. Multiple OSDs may share the same
647 journal disk (using separate partitions); this way, a single fast SSD
648 can serve as journal for multiple spinning disks.
650 The second use case currently requires parted, so a Recommends: for
651 parted has been added to Debian packaging.
655 Signed-off-by: Tommi Virtanen <tv@inktank.com>
657 commit 5a3076fd51706e32bc55b734099037520299003b
658 Author: Yehuda Sadeh <yehuda@inktank.com>
659 Date: Mon Oct 15 09:43:47 2012 -0700
661 rgw: don't add port to url if already has one
664 Specifically, is host name string already has ':', then
665 don't try to append theport (swift auth).
668 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
670 commit d76b71caa99a39035ea6ed9db6ca0994975d38af
671 Author: Sage Weil <sage@inktank.com>
672 Date: Mon Oct 15 16:37:05 2012 -0700
674 admin_socket: fix '0' protocol version
676 Broken by 895e24d198ced83ab7fed3725f12f75e3bc97b0b.
678 Signed-off-by: Sage Weil <sage@inktank.com>
680 commit 714254cb0522b194118eb854f90d4c3971e8c6b3
681 Author: Sage Weil <sage@inktank.com>
682 Date: Mon Oct 8 17:14:22 2012 -0700
684 mon: drop command replies on paxos reset
686 If paxos resets, do not send the reply for the commit we were waiting for;
687 let the command be reprocessed and re-proposed.
689 Among other things, this could lead to nondeterministic results for
690 'ceph osd create <uuid>'.
692 Signed-off-by: Sage Weil <sage@inktank.com>
694 commit d9e981f4209da9a0e000d434e28686a133b25f11
695 Merge: 5d84b4a 13b448b
696 Author: Sage Weil <sage@inktank.com>
697 Date: Mon Oct 8 21:02:51 2012 -0700
699 Merge remote-tracking branch 'gh/for-stable-fstypes-and-ext-journal' into stable
701 commit 5d84b4a22b2ff975679711719f336a312a6edff3
702 Author: Tommi Virtanen <tv@inktank.com>
703 Date: Thu Aug 2 13:02:04 2012 -0700
705 ceph-authtool: Fix usage, it's --print-key not --print.
707 Signed-off-by: Tommi Virtanen <tv@inktank.com>
709 commit 13b448b8ce959f31d07cd1827614c445c410aa62
710 Author: Tommi Virtanen <tv@inktank.com>
711 Date: Fri Oct 5 09:22:34 2012 -0700
713 upstart: OSD journal can be a symlink; if it's dangling, don't start.
715 This lets a $osd_data/journal symlink point to
716 /dev/disk/by-partuuid/UUID and the osd will not attempt to start until
717 that disk is available.
719 Signed-off-by: Tommi Virtanen <tv@inktank.com>
721 commit d9dfa2d17859ca07d735b8e35c5dc00a25995153
722 Author: Sage Weil <sage@inktank.com>
723 Date: Fri Oct 5 09:10:31 2012 -0700
725 osd: Make --get-journal-fsid not really start the osd.
727 This way, it won't need -i ID and it won't access the osd_data_dir.
728 That makes it useful for locating the right osd to use with an
729 external journal partition.
731 Signed-off-by: Tommi Virtanen <tv@inktank.com>
733 commit f65daf3f24830a28e85ebe7c6fce2642f5fc6dd8
734 Author: Tommi Virtanen <tv@inktank.com>
735 Date: Fri Oct 5 09:08:56 2012 -0700
737 osd: Make --get-journal-fsid not attempt aio or direct_io.
739 Signed-off-by: Tommi Virtanen <tv@inktank.com>
741 commit 86546d7dec6cd5080162d1f3d7de2c579736dd65
742 Author: Tommi Virtanen <tv@inktank.com>
743 Date: Thu Oct 4 16:03:40 2012 -0700
745 ceph-disk-prepare: Use the OSD uuid as the partition GUID.
747 This will make locating the right data partition for a given journal
748 partition a lot easier.
750 Signed-off-by: Tommi Virtanen <tv@inktank.com>
752 commit fa3aa41ead0be5a27695e3c208c5af00af90bc53
753 Author: Tommi Virtanen <tv@inktank.com>
754 Date: Wed Oct 3 12:38:38 2012 -0700
756 debian/control, ceph-disk-prepare: Depend on xfsprogs, use xfs by default.
758 Ext4 as a default is a bad choice, as we don't perform enough QA with
759 it. To use XFS as the default for ceph-disk-prepare, we need to depend
762 btrfs-tools is already recommended, so no change there. If you set
763 osd_fs_type=btrfs, and don't have the package installed, you'll just
764 get an error message.
766 Signed-off-by: Tommi Virtanen <tv@inktank.com>
768 commit 4c3997dc8bbd3985a7b86d8c54b71fbe15af1591
769 Author: Tommi Virtanen <tv@inktank.com>
770 Date: Wed Oct 3 10:13:17 2012 -0700
772 ceph-disk-{prepare,activate}: Default mkfs arguments and mount options.
774 The values for the settings were copied from teuthology task "ceph".
776 Signed-off-by: Tommi Virtanen <tv@inktank.com>
778 commit a6229b901785ef0a88d1b1886b5b5ae339ac7fcb
779 Author: Tommi Virtanen <tv@inktank.com>
780 Date: Wed Oct 3 08:47:20 2012 -0700
782 ceph-disk-prepare: Avoid triggering activate before prepare is done.
784 Earlier testing never saw this, but now a mount of a disk triggers a
785 udev blockdev-added event, causing ceph-disk-activate to run even
786 before ceph-disk-prepare has had a chance to write the files and
789 Avoid this by using a temporary partition type uuid ("ceph 2 be"), and
790 only setting it to the permanent ("ceph osd"). The hotplug event won't
791 match the type uuid, and thus won't trigger ceph-disk-activate.
793 Signed-off-by: Tommi Virtanen <tv@inktank.com>
795 commit 03cec77e2b0b4536349880de6890ad9537da0783
796 Author: Tommi Virtanen <tv@inktank.com>
797 Date: Tue Oct 2 17:06:11 2012 -0700
799 ceph-disk-activate: Add a comment about user_xattr being default now.
801 Signed-off-by: Tommi Virtanen <tv@inktank.com>
803 commit 4aa13c6d228b9256870e02a57496f00a75e17349
804 Author: Tommi Virtanen <tv@inktank.com>
805 Date: Tue Oct 2 16:53:35 2012 -0700
807 ceph-disk-activate: Use mount options from ceph.conf
809 Always uses default cluster name ("ceph") for now, see
810 http://tracker.newdream.net/issues/3253
813 Signed-off-by: Tommi Virtanen <tv@inktank.com>
815 commit b684ab1d1727b23f75f232123f1da947df2e5e63
816 Author: Tommi Virtanen <tv@inktank.com>
817 Date: Tue Oct 2 16:43:08 2012 -0700
819 ceph-disk-activate: Refactor to extract detect_fstype call.
821 This allows us to use the fstype for a config lookup.
823 Signed-off-by: Tommi Virtanen <tv@inktank.com>
825 commit ce5e1f8dfaa8675b0a1e9fa22ebf91be9652aa01
826 Author: Tommi Virtanen <tv@inktank.com>
827 Date: Tue Oct 2 16:37:07 2012 -0700
829 ceph-disk-activate: Unmount on errors (if it did the mount).
831 This cleans up the error handling to not leave disks mounted
832 in /var/lib/ceph/tmp/mnt.* when something fails, e.g. when
833 the ceph command line tool can't talk to mons.
835 Signed-off-by: Tommi Virtanen <tv@inktank.com>
837 commit da86e4e43676ceaf0b39635954489c297de88025
838 Author: Tommi Virtanen <tv@inktank.com>
839 Date: Tue Oct 2 16:23:55 2012 -0700
841 ceph-disk-prepare: Allow setting mkfs arguments and mount options in ceph.conf
843 Tested with meaningless but easy-to-verify values:
847 osd_fs_mkfs_arguments_xfs = -i size=512
848 osd_fs_mount_options_xfs = noikeep
850 ceph-disk-activate does not respect the mount options yet.
853 Signed-off-by: Tommi Virtanen <tv@inktank.com>
855 commit cf2e3b15cea99cc100169e7a67103676d8990aee
856 Author: Tommi Virtanen <tv@inktank.com>
857 Date: Tue Oct 2 16:04:15 2012 -0700
859 ceph-disk-prepare: Allow specifying fs type to use.
861 Either use ceph.conf variable osd_fs_type or command line option
864 Default is still ext4, as currently nothing guarantees xfsprogs
865 or btrfs-tools are installed.
867 Currently both btrfs and xfs seems to trigger a disk hotplug event at
868 mount time, thus triggering a useless and unwanted ceph-disk-activate
869 run. This will be worked around in a later commit.
871 Currently mkfs and mount options cannot be configured.
874 Signed-off-by: Tommi Virtanen <tv@inktank.com>
876 commit f031046c4d600ce0a0c2c9ad5856dc100b0901fb
877 Author: Yehuda Sadeh <yehuda@inktank.com>
878 Date: Wed Sep 26 15:43:56 2012 -0700
880 rgw: copy_object should not override ETAG implicitly
882 When copying an object with new attrs, we still need to
885 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
887 commit 2fddfb0a95288d39cee2e82376297ce43296743b
888 Author: Yehuda Sadeh <yehuda@inktank.com>
889 Date: Mon Sep 24 18:10:24 2012 -0700
891 rgw: url_decode should allocate extra byte for dest
893 Was missing extra byte for null termination
895 Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>