ceph/doc/changelog/v0.61.5.txt

   1 commit 8ee10dc4bb73bdd918873f29c70eedc3c7ef1979
   2 Author: Gary Lowell <gary.lowell@inktank.com>
   3 Date:   Wed Jul 17 16:39:08 2013 -0700
   4
   5     v0.61.5
   6
   7 commit 39bffac6b6c898882d03de392f7f2218933d942b
   8 Author: Sage Weil <sage@inktank.com>
   9 Date:   Tue Jul 16 13:14:50 2013 -0700
  10
  11     ceph-disk: rely on /dev/disk/by-partuuid instead of special-casing journal symlinks
  12
  13     This was necessary when ceph-disk-udev didn't create the by-partuuid (and
  14     other) symlinks for us, but now it is fragile and error-prone.  (It also
  15     appears to be broken on a certain customer RHEL VM.)  See
  16     d7f7d613512fe39ec883e11d201793c75ee05db1.
  17
  18     Instead, just use the by-partuuid symlinks that we spent all that ugly
  19     effort generating.
  20
  21     Backport: cuttlefish
  22     Signed-off-by: Sage Weil <sage@inktank.com>
  23     Reviewed-by: Dan Mick <dan.mick@inktank.com>
  24     (cherry picked from commit 64379e701b3ed862c05f156539506d3382f77aa8)
  25
  26 commit 363d54288254b5e2311cd28fce5988d68cfd5773
  27 Author: Joao Eduardo Luis <joao.luis@inktank.com>
  28 Date:   Tue Jul 16 16:49:48 2013 +0100
  29
  30     mon: Monitor: StoreConverter: clearer debug message on 'needs_conversion()'
  31
  32     The previous debug message outputted the function's name, as often our
  33     functions do.  This was however a source of bewilderment, as users would
  34     see those in logs and think their stores would need conversion.  Changing
  35     this message is trivial enough and it will make ceph users happier log
  36     readers.
  37
  38     Backport: cuttlefish
  39     Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
  40     Reviewed-by: Sage Weil <sage@inktank.com>
  41     (cherry picked from commit ad1392f68170b391d11df0ce5523c2d1fb57f60e)
  42
  43 commit 0ea89760def73f76d8100889eca3c25b0a6eb772
  44 Author: Joao Eduardo Luis <joao.luis@inktank.com>
  45 Date:   Tue Jul 16 16:45:39 2013 +0100
  46
  47     mon: Monitor: do not reopen MonitorDBStore during conversion
  48
  49     We already open the store on ceph_mon.cc, before we start the conversion.
  50     Given we are unable to reproduce this every time a conversion is triggered,
  51     we are led to believe that this causes a race in leveldb that will lead
  52     to 'store.db/LOCK' being locked upon the open this patch removes.
  53
  54     Regardless, reopening the db here is pointless as we already did it when
  55     we reach Monitor::StoreConverter::convert().
  56
  57     Fixes: #5640
  58     Backport: cuttlefish
  59
  60     Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
  61     Reviewed-by: Sage Weil <sage@inktank.com>
  62     (cherry picked from commit 036e6739a4e873863bae3d7d00f310c015dfcdb3)
  63
  64 commit 595c09df9134fb0d62144fe1594914c90e567dca
  65 Author: Sage Weil <sage@inktank.com>
  66 Date:   Sun Jul 14 15:54:29 2013 -0700
  67
  68     messages/MClientReconnect: clear data when encoding
  69
  70     The MClientReconnect puts everything in the data payload portion of
  71     the message and nothing in the front portion.  That means that if the
  72     message is resent (socket failure or something), the messenger thinks it
  73     hasn't been encoded yet (front empty) and reencodes, which means
  74     everything gets added (again) to the data portion.
  75
  76     Decoding keep decoding until it runs out of data, so the second copy
  77     means we decode garbage snap realms, leading to the crash in bug
  78
  79     Clearing data each time around resolves the problem, although it does
  80     mean we do the encoding work multiple times.  We could alternatively
  81     (or also) stick some data in the front portion of the payload
  82     (ignored), but that changes the wire protocol and I would rather not
  83     do that.
  84
  85     Fixes: #4565
  86     Backport: cuttlefish
  87     Signed-off-by: Sage Weil <sage@inktank.com>
  88     Reviewed-by: Greg Farnum <greg@inktank.com>
  89     (cherry picked from commit 314cf046b0b787ca69665e8751eab6fe7adb4037)
  90
  91 commit 8c178e0d39d8d4a4820eb061f79d74f95e60199f
  92 Author: Sage Weil <sage@inktank.com>
  93 Date:   Mon Jul 15 10:05:37 2013 -0700
  94
  95     mon: once sync full is chosen, make sure we don't change our mind
  96
  97     It is possible for a sequence like:
  98
  99      - probe
 100      - first probe reply has paxos trim that indicates a full sync is
 101        needed
 102      - start sync
 103      - clear store
 104      - something happens that makes us abort and bootstrap (e.g., the
 105        provider mon restarts
 106      - probe
 107      - first probe reply has older paxos trim bound and we call an election
 108      - on election completion, we crash because we have no data.
 109
 110     Non-determinism of the probe decision aside, we need to ensure that
 111     the info we share during probe (fc, lc) is accurate, and that once we
 112     clear the store we know we *must* do a full sync.
 113
 114     This is a backport of aa60f940ec1994a61624345586dc70d261688456.
 115
 116     Fixes: #5621
 117     Signed-off-by: Sage Weil <sage@inktank.com>
 118     Reviewed-by: Greg Farnum <greg@inktank.com>
 119
 120 commit 6af0ed9bc4cc955f8c30ad9dc6e9095599f323d0
 121 Author: Sage Weil <sage@inktank.com>
 122 Date:   Tue Jul 9 14:12:15 2013 -0700
 123
 124     mon: do not scrub if scrub is in progress
 125
 126     This prevents an assert from unexpected scrub results from the previous
 127     scrub on the leader.
 128
 129     Backport: cuttlefish
 130     Signed-off-by: Sage Weil <sage@inktank.com>
 131     (cherry picked from commit 00ae543b3e32f89d906a0e934792cc5309f57696)
 132
 133 commit 5642155ba5ca9b384a7af058a6538ac00c2a592d
 134 Author: Sage Weil <sage@inktank.com>
 135 Date:   Wed Jul 10 10:06:20 2013 -0700
 136
 137     messages/MPGStats: do not set paxos version to osdmap epoch
 138
 139     The PaxosServiceMessage version field is meant for client-coordinated
 140     ordering of messages when switching between monitors (and is rarely
 141     used).  Do not fill it with the osdmap epoch lest it be compared to a
 142     pgmap version, which may cause the mon to (near) indefinitely put it on
 143     a wait queue until the pgmap version catches up.
 144
 145     Backport: cuttlefish
 146     Signed-off-by: Sage Weil <sage@inktank.com>
 147     Reviewed-by: Greg Farnum <greg@inktank.com>
 148     (cherry picked from commit b36338be43f43b6dd4ee87c97f2eaa23b467c386)
 149
 150 commit 06c65988bb0b1d1ec626fe31e9d806a1c4e24b28
 151 Author: Sage Weil <sage@inktank.com>
 152 Date:   Thu Jul 11 18:43:24 2013 -0700
 153
 154     osd/OSDmap: fix OSDMap::Incremental::dump() for new pool names
 155
 156     The name is always present when pools are created, but not when they are
 157     modified.  Also, a name may be present with a new_pools entry if the pool
 158     is just renamed.  Separate it out completely in the dump.
 159
 160     Backport: cuttlefish, bobtail
 161     Signed-off-by: Sage Weil <sage@inktank.com>
 162     (cherry picked from commit 3e4a29111e89588385e63f8d92ce3d67739dd679)
 163
 164 commit 658240710baaf9c661b8fbf856322907a0d394ee
 165 Author: Sage Weil <sage@inktank.com>
 166 Date:   Mon Jul 8 10:49:28 2013 -0700
 167
 168     mon/PaxosService: prevent reads until initial service commit is done
 169
 170     Do not process reads (or, by PaxosService::dispatch() implication, writes)
 171     until we have committed the initial service state.  This avoids things like
 172     EPERM due to missing keys when we race with mon creation, triggered by
 173     teuthology tests doing their health check after startup.
 174
 175     Fixes: #5515
 176     Backport: cuttlefish
 177     Signed-off-by: Sage Weil <sage@inktank.com>
 178     Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
 179     (cherry picked from commit d08b6d6df7dba06dad73bdec2c945f24afc02717)
 180
 181 commit 5c3ff33771e227b3fb5cc354323846fe8db4ecc1
 182 Author: Sage Weil <sage@inktank.com>
 183 Date:   Fri Jun 28 12:21:58 2013 -0700
 184
 185     client: send all request put's through put_request()
 186
 187     Make sure all MetaRequest reference put's go through the same path that
 188     releases inode references, including all of the error paths.
 189
 190     Signed-off-by: Sage Weil <sage@inktank.com>
 191     (cherry picked from commit 87217e1e3cb2785b79d0dec49bd3f23a827551f5)
 192
 193 commit 1df78ad73df581bc7537688ae28bda820b089a13
 194 Author: Sage Weil <sage@inktank.com>
 195 Date:   Fri Jun 28 11:50:11 2013 -0700
 196
 197     client: fix remaining Inode::put() caller, and make method psuedo-private
 198
 199     Not sure I can make this actually private and make Client::put_inode() a
 200     friend method (making all of Client a friend would defeat the purpose).
 201     This works well enough, though!
 202
 203     Signed-off-by: Sage Weil <sage@inktank.com>
 204     (cherry picked from commit 9af3b86b25574e4d2cdfd43e61028cffa19bdeb1)
 205
 206 commit fea024cc3dd2c6fd9ff322d1cd15e0d75c92eca5
 207 Author: Sage Weil <sage@inktank.com>
 208 Date:   Thu Jun 27 21:39:35 2013 -0700
 209
 210     client: use put_inode on MetaRequest inode refs
 211
 212     When we drop the request inode refs, we need to use put_inode() to ensure
 213     they get cleaned up properly (removed from inode_map, caps released, etc.).
 214     Do this explicitly here (as we do with all other inode put() paths that
 215     matter).
 216
 217     Fixes: #5381
 218     Backport: cuttlefish
 219     Signed-off-by: Sage Weil <sage@inktank.com>
 220     (cherry picked from commit 81bee6487fb1ce9e090b030d61bda128a3cf4982)
 221
 222 commit 62ae39ec8f208cb8f89e43ba844b9a20b4315c61
 223 Author: Sage Weil <sage@inktank.com>
 224 Date:   Mon Jul 8 15:57:48 2013 -0700
 225
 226     mon: be smarter about calculating last_epoch_clean lower bound
 227
 228     We need to take PGs whose mapping has not changed in a long time into
 229     account.  For them, the pg state will indicate it was clean at the time of
 230     the report, in which case we can use that as a lower-bound on their actual
 231     latest epoch clean.  If they are not currently clean (at report time), use
 232     the last_epoch_clean value.
 233
 234     Fixes: #5519
 235     Signed-off-by: Sage Weil <sage@inktank.com>
 236     (cherry picked from commit cc0006deee3153e06ddd220bf8a40358ba830135)
 237
 238 commit da725852190245d2f91b7b21e72baee70e4342bd
 239 Author: Sage Weil <sage@inktank.com>
 240 Date:   Mon Jul 8 13:27:58 2013 -0700
 241
 242     osd: report pg stats to mon at least every N (=500) epochs
 243
 244     The mon needs a moderately accurate last_epoch_clean value in order to trim
 245     old osdmaps.  To prevent a PG that hasn't peered or received IO in forever
 246     from preventing this, send pg stats at some minimum frequency.  This will
 247     increase the pg stat report workload for the mon over an idle pool, but
 248     should be no worse that a cluster that is getting actual IO and sees these
 249     updates from normal stat updates.
 250
 251     This makes the reported update a bit more aggressive/useful in that the epoch
 252     is the last map epoch processed by this PG and not just one that is >= the
 253     currenting interval.  Note that the semantics of this field are pretty useless
 254     at this point.
 255
 256     See #5519
 257
 258     Signed-off-by: Sage Weil <sage@inktank.com>
 259     (cherry picked from commit da81228cc73c95737f26c630e5c3eccf6ae1aaec)
 260
 261 commit 757af91b2af0da6bbfeeb53551fa1ef4ef9118ea
 262 Author: Sage Weil <sage@inktank.com>
 263 Date:   Wed Jul 10 11:32:34 2013 -0700
 264
 265     osd: fix warning
 266
 267     From 653e04a79430317e275dd77a46c2b17c788b860b
 268
 269     Backport: cuttlefish, bobtail
 270     Signed-off-by: Sage Weil <sage@inktank.com>
 271     (cherry picked from commit bc291d3fc3fc1cac838565cbe0f25f71d855a6e3)
 272
 273 commit 65af2538329472d2fd078bb961863c40cdabda12
 274 Merge: e537699 804314b
 275 Author: Sage Weil <sage@inktank.com>
 276 Date:   Fri Jul 12 15:21:20 2013 -0700
 277
 278     Merge remote-tracking branch 'gh/wip-mon-sync-2' into cuttlefish
 279
 280     Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
 281     Reviewed-by: Greg Farnum <greg@inktank.com>
 282
 283 commit e537699b33f84c14f027b56372fbcb0a99bbe88d
 284 Author: Sandon Van Ness <sandon@inktank.com>
 285 Date:   Wed Jul 10 14:55:52 2013 -0700
 286
 287     Get device-by-path by looking for it instead of assuming 3rd entry.
 288
 289     On some systems (virtual machines so far) the device-by-path entry
 290     from udevadm is not always in the same spot so instead actually
 291     look for the right output instead of blindy assuming that its a
 292     specific field in the output.
 293
 294     Signed-off-by: Sandon Van Ness <sandon@inktank.com>
 295     Reviewed-by: Gary Lowell  <gary.lowell@inktank.com>
 296
 297 commit 804314b8bfa5ec75cc9653e2928874c457395c92
 298 Merge: 6ad9fe1 78f2266
 299 Author: Sage Weil <sage@inktank.com>
 300 Date:   Wed Jul 10 11:40:37 2013 -0700
 301
 302     Merge remote-tracking branch 'gh/cuttlefish' into wip-mon-sync-2
 303
 304 commit 78f226634bd80f6678b1f74ccf785bc52fcd6b62
 305 Author: Sage Weil <sage@inktank.com>
 306 Date:   Wed Jul 10 11:02:08 2013 -0700
 307
 308     osd: limit number of inc osdmaps send to peers, clients
 309
 310     We should not send an unbounded number of inc maps to our peers or clients.
 311     In particular, if a peer is not contacted for a while, we may think they
 312     have a very old map (say, 10000 epochs ago) and send thousands of inc maps
 313     when the distribution shifts and we need to peer.
 314
 315     Note that if we do not send enough maps, the peers will make do by
 316     requesting the map from somewhere else (currently the mon).  Regardless
 317     of the source, however, we must limit the amount that we speculatively
 318     share as it usually is not needed.
 319
 320     Backport: cuttlefish, bobtail
 321     Signed-off-by: Sage Weil <sage@inktank.com>
 322     Reviewed-by: Samuel Just <sam.just@inktank.com>
 323     (cherry picked from commit 653e04a79430317e275dd77a46c2b17c788b860b)
 324
 325 commit 54ee2dc80ed032c286546da51442340ec9991cdf
 326 Author: Christophe Courtaut <christophe.courtaut@gmail.com>
 327 Date:   Mon Jul 1 14:57:17 2013 +0200
 328
 329     rgw: Fix return value for swift user not found
 330
 331     http://tracker.ceph.com/issues/1779 fixes #1779
 332
 333     Adjust the return value from rgw_get_user_info_by_swift call
 334     in RGW_SWIFT_Auth_Get::execute() to have the correct
 335     return code in response.
 336     (cherry picked from commit 4089001de1f22d6acd0b9f09996b71c716235551)
 337
 338 commit 47852c263831707fff1570317a7446b0700c5962
 339 Author: Sage Weil <sage@inktank.com>
 340 Date:   Tue Jul 9 21:55:51 2013 -0700
 341
 342     mon/OSDMonitor: make 'osd crush rm ...' slightly more idempotent
 343
 344     This is a manual backport of 18a624fd8b90d9959de51f07622cf0839e6bd9aa.
 345     Do not return immediately if we are looking at uncommitted state.t
 346
 347     Signed-off-by: Sage Weil <sage@inktank.com>
 348
 349 commit bfc26c656d183fbcc90a352391e47f9f51c96052
 350 Author: Sage Weil <sage@inktank.com>
 351 Date:   Mon Jul 8 17:46:40 2013 -0700
 352
 353     mon/OSDMonitor: fix base case for loading full osdmap
 354
 355     Right after cluster creation, first_committed is 1 and latest stashed in 0,
 356     but we don't have the initial full map yet.  Thereafter, we do (because we
 357     write it with trim).  Fixes afd6c7d8247075003e5be439ad59976c3d123218.
 358
 359     Signed-off-by: Sage Weil <sage@inktank.com>
 360     Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
 361     (cherry picked from commit 43fa7aabf1f7e5deb844c1f52d451bab9e7d1006)
 362
 363 commit 7fb3804fb860dcd0340dd3f7c39eec4315f8e4b6
 364 Author: Sage Weil <sage@inktank.com>
 365 Date:   Mon Jul 8 15:04:59 2013 -0700
 366
 367     mon: fix osdmap stash, trim to retain complete history of full maps
 368
 369     The current interaction between sync and stashing full osdmaps only on
 370     active mons means that a sync can result in an incomplete osdmap_full
 371     history:
 372
 373      - mon.c starts a full sync
 374      - during sync, active osdmap service should_stash_full() is true and
 375        includes a full in the txn
 376      - mon.c sync finishes
 377      - mon.c update_from_paxos gets "latest" stashed that it got from the
 378        paxos txn
 379      - mon.c does *not* walk to previous inc maps to complete it's collection
 380        of full maps.
 381
 382     To fix this, we disable the periodic/random stash of full maps by the
 383     osdmap service.
 384
 385     This introduces a new problem: we must have at least one full map (the first
 386     one) in order for a mon that just synced to build it's full collection.
 387     Extend the encode_trim() process to allow the osdmap service to include
 388     the oldest full map with the trim txn.  This is more complex than just
 389     writing the full maps in the txn, but cheaper--we only write the full
 390     map at trim time.
 391
 392     This *might* be related to previous bugs where the full osdmap was
 393     missing, or case where leveldb keys seemed to 'disappear'.
 394
 395     Fixes: #5512
 396     Backport: cuttlefish
 397     Signed-off-by: Sage Weil <sage@inktank.com>
 398     Reviewed-by: Greg Farnum <greg@inktank.com>
 399     (cherry picked from commit afd6c7d8247075003e5be439ad59976c3d123218)
 400
 401 commit 24f90b832c695ef13021db66a178c18369ac356d
 402 Author: Sage Weil <sage@inktank.com>
 403 Date:   Mon Jul 8 15:07:57 2013 -0700
 404
 405     mon: implement simple 'scrub' command
 406
 407     Compare all keys within the sync'ed prefixes across members of the quorum
 408     and compare the key counts and CRC for inconsistencies.
 409
 410     Currently this is a one-shot inefficient hammer.  We'll want to make this
 411     work in chunks before it is usable in production environments.
 412
 413     Protect with a feature bit to avoid sending MMonScrub to mons who can't
 414     decode it.
 415
 416     Signed-off-by: Sage Weil <sage@inktank.com>
 417     Reviewed-by: Greg Farnum <greg@inktank.com>
 418     (cherry picked from commit a9906641a1dce150203b72682da05651e4d68ff5)
 419
 420     Conflicts:
 421
 422         src/mon/MonCommands.h
 423         src/mon/Monitor.cc
 424
 425 commit 926f723c12428a034545c6c4ff6641e1d5e05d24
 426 Author: Samuel Just <sam.just@inktank.com>
 427 Date:   Wed Jul 3 11:18:33 2013 -0700
 428
 429     Elector.h: features are 64 bit
 430
 431     Fixes: #5497
 432     Signed-off-by: Samuel Just <sam.just@inktank.com>
 433     Reviewed-by: Sage Weil <sage@inktank.com>
 434     Reviewed-by: Joao Luis <joao.luis@inktank.com>
 435     (cherry picked from commit 3564e304e3f50642e4d9ff25e529d5fc60629093)
 436
 437 commit c2b38291e706c9d1d4d337cee3a944f34bf66525
 438 Author: Samuel Just <sam.just@inktank.com>
 439 Date:   Wed Jul 3 11:18:19 2013 -0700
 440
 441     ceph_features.h: declare all features as ULL
 442
 443     Otherwise, the first 32 get |'d together as ints.  Then, the result
 444     ((int)-1) is sign extended to ((long long int)-1) before being |'d
 445     with the 1LL entries.  This results in ~((uint64_t)0).
 446
 447     Fixes: #5497
 448     Signed-off-by: Samuel Just <sam.just@inktank.com>
 449     Reviewed-by: Sage Weil <sage@inktank.com>
 450     Reviewed-by: Joao Luis <joao.luis@inktank.com>
 451     (cherry picked from commit 4255b5c2fb54ae40c53284b3ab700fdfc7e61748)
 452
 453 commit 95ef961d8537fc369efd0634262ffb8f288d6e9e
 454 Author: Samuel Just <sam.just@inktank.com>
 455 Date:   Tue Jul 2 21:09:36 2013 -0700
 456
 457     Pipe: use uint64_t not unsigned when setting features
 458
 459     Fixes: #5497
 460     Signed-off-by: Samuel Just <sam.just@inktank.com>
 461     Reviewed-by: Sage Weil <sage@inktank.com>
 462     Reviewed-by: Joao Luis <joao.luis@inktank.com>
 463     (cherry picked from commit bc3e2f09f8860555d8b3b49b2eea164b4118d817)
 464
 465 commit 09d258b70a28e5cea555b9d7e215fe41d6b84577
 466 Author: Sage Weil <sage@inktank.com>
 467 Date:   Mon Jul 8 11:24:48 2013 -0700
 468
 469     client: remove O_LAZY
 470
 471     The once-upon-a-time unique O_LAZY value I chose forever ago is now
 472     O_NOATIME, which means that some clients are choosing relaxed
 473     consistency without meaning to.
 474
 475     It is highly unlikely that a real O_LAZY will ever exist, and we can
 476     select it in the ceph case with the ioctl or libcephfs call, so drop
 477     any support for doing this via open(2) flags.
 478
 479     Update doc/lazy_posix.txt file re: lazy io.
 480
 481     Backport: cuttlefish
 482     Signed-off-by: Sage Weil <sage@inktank.com>
 483     Reviewed-by: Greg Farnum <greg@inktank.com>
 484     (cherry picked from commit 94afedf02d07ad4678222aa66289a74b87768810)
 485
 486 commit c3b684932bad31fc853ad556d16e1e4a9926486e
 487 Author: Sage Weil <sage@inktank.com>
 488 Date:   Mon Jul 8 12:55:20 2013 -0700
 489
 490     osd/osd_types: fix pg_stat_t::dump for last_epoch_clean
 491
 492     Backport: cuttlefish
 493     Signed-off-by: Sage Weil <sage@inktank.com>
 494     (cherry picked from commit 69a55445439fce0dd6a3d32ff4bf436da42f1b11)
 495
 496 commit a02f2510fcc800b9f2cf2a06401a7b97d5985409
 497 Author: Sage Weil <sage@inktank.com>
 498 Date:   Fri Jul 5 16:03:49 2013 -0700
 499
 500     mon: remove bad assert about monmap version
 501
 502     It is possible to start a sync when our newest monmap is 0.  Usually we see
 503     e0 from probe, but that isn't always published as part of the very first
 504     paxos transaction due to the way PaxosService::_active generates it's
 505     first initial commit.
 506
 507     In any case, having e0 here is harmless.
 508
 509     Fixes: #5509
 510     Signed-off-by: Sage Weil <sage@inktank.com>
 511     Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
 512     (cherry picked from commit 85a1d6cc5d3852c94d1287b566656c5b5024fa13)
 513
 514 commit 6ad9fe17a674ba65bbeb4052cb1ac47f3113e7bf
 515 Author: Sage Weil <sage@inktank.com>
 516 Date:   Thu Jul 4 19:33:06 2013 -0700
 517
 518     mon/Paxos: fix sync restart
 519
 520     If we have a sync going, and an election intervenes, the client will
 521     try to continue by sending a new start_chunks request.  In order to
 522     ensure that we get all of the paxos commits from our original starting
 523     point (and thus properly update the keys from which they started),
 524     only pay attention if they *also* send their current last_committed
 525     version.  Otherwise, start them at the beginning.
 526
 527     Signed-off-by: Sage Weil <sage@inktank.com>
 528
 529 commit c5812b1c893305a7d20f9eaec2695c8b1691f0c9
 530 Author: Sage Weil <sage@inktank.com>
 531 Date:   Thu Jul 4 14:57:06 2013 -0700
 532
 533     mon: uninline _trim_enable and Paxos::trim_{enable,disable} so we can debug them
 534
 535     Signed-off-by: Sage Weil <sage@inktank.com>
 536
 537 commit 6fbcbd7fddf35a5be4b38e536871903bff4f9bf1
 538 Author: Sage Weil <sage@inktank.com>
 539 Date:   Thu Jul 4 14:55:34 2013 -0700
 540
 541     mon/Paxos: increase paxos max join drift
 542
 543     A value of 10 is too aggressive for large, long-running syncs. 100 is
 544     about 2 minutes of activity at most, which should be a more forgiving
 545     buffer.
 546
 547     Signed-off-by: Sage Weil <sage@inktank.com>
 548
 549 commit f3a51fa30e5ce1656853b40d831409f195f6e4ca
 550 Author: Sage Weil <sage@inktank.com>
 551 Date:   Thu Jul 4 14:21:04 2013 -0700
 552
 553     mon/Paxos: configure minimum paxos txns separately
 554
 555     We were using paxos_max_join_drift to control the minimum number of
 556     paxos transactions to keep around.  Instead, make this explicit, and
 557     separate from the join drift.
 558
 559     Signed-off-by: Sage Weil <sage@inktank.com>
 560
 561 commit 1156721f22f5f337241eef3d0276ca74fe6352d1
 562 Author: Sage Weil <sage@inktank.com>
 563 Date:   Thu Jul 4 17:09:07 2013 -0700
 564
 565     mon: include any new paxos commits in each sync CHUNK message
 566
 567     We already take note of the paxos version when we begin the sync.  As
 568     sync progresses and there are new paxos commits/txns, include those
 569     and update last_committed, so that when sync completes we will have
 570     a full view of everything that happened during sync.
 571
 572     Note that this does not introduce any compatibility change.  This change
 573     *only* affects the provider.  The key difference is that at the end
 574     of the sync, the provide will set version to the latest version, and
 575     not the version from the start of the sync (as was done previously).
 576
 577     Signed-off-by: Sage Weil <sage@inktank.com>
 578
 579 commit 40672219a081f0dc2dd536977290ef05cfc9f097
 580 Author: Sage Weil <sage@inktank.com>
 581 Date:   Thu Jul 4 12:17:28 2013 -0700
 582
 583     mon/MonitorDBStore: expose get_chunk_tx()
 584
 585     Allow users get the transaction unencoded.
 586
 587     Signed-off-by: Sage Weil <sage@inktank.com>
 588
 589 commit db2bb270e93ed44f9252d65d1d4c9b36875d0ea5
 590 Author: Sage Weil <sage@inktank.com>
 591 Date:   Wed Jul 3 17:15:56 2013 -0700
 592
 593     mon: enable leveldb cache by default
 594
 595     256 is not as large as the upstream 512 MB, but will help signficiantly and
 596     be less disruptive for existing cuttlefish clusters.
 597
 598     Sort-of backport of e93730b7ffa48b53c8da2f439a60cb6805facf5a.
 599
 600     Signed-off-by: Sage Weil <sage@inktank.com>
 601
 602 commit 123f676e3ae8154ca94cb076c4c4ec5389d2a643
 603 Author: Sage Weil <sage@inktank.com>
 604 Date:   Wed Jul 3 16:56:06 2013 -0700
 605
 606     mon/Paxos: make 'paxos trim disabled max versions' much much larger
 607
 608     108000 is about 3 hours if paxos is going full-bore (1 proposal/second).
 609     That ought to be pretty safe.  Otherwise, we start trimming to soon and a
 610     slow sync will just have to restart when it finishes.
 611
 612     Backport: cuttlefish
 613     Signed-off-by: Sage Weil <sage@inktank.com>
 614     Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
 615     (cherry picked from commit 71ebfe7e1abe4795b46cf00dfe1b03d1893368b0)
 616
 617     Conflicts:
 618
 619         src/common/config_opts.h
 620
 621 commit 03393c0df9f54e4f1db60e1058ca5a7cd89f44e6
 622 Author: Sage Weil <sage@inktank.com>
 623 Date:   Wed Jun 26 06:01:40 2013 -0700
 624
 625     mon: do not reopen MonitorDBStore during startup
 626
 627     level doesn't seem to like this when it races with an internal compaction
 628     attempt (see below).  Instead, let the store get opened by the ceph_mon
 629     caller, and pull a bit of the logic into the caller to make the flow a
 630     little easier to follow.
 631
 632         -2> 2013-06-25 17:49:25.184490 7f4d439f8780 10 needs_conversion
 633         -1> 2013-06-25 17:49:25.184495 7f4d4065c700  5 asok(0x13b1460) entry start
 634          0> 2013-06-25 17:49:25.316908 7f4d3fe5b700 -1 *** Caught signal (Segmentation fault) **
 635      in thread 7f4d3fe5b700
 636
 637      ceph version 0.64-667-g089cba8 (089cba8fc0e8ae8aef9a3111cba7342ecd0f8314)
 638      1: ceph-mon() [0x649f0a]
 639      2: (()+0xfcb0) [0x7f4d435dccb0]
 640      3: (leveldb::Table::BlockReader(void*, leveldb::ReadOptions const&, leveldb::Slice const&)+0x154) [0x806e54]
 641      4: ceph-mon() [0x808840]
 642      5: ceph-mon() [0x808b39]
 643      6: ceph-mon() [0x806540]
 644      7: (leveldb::DBImpl::DoCompactionWork(leveldb::DBImpl::CompactionState*)+0xdd) [0x7f363d]
 645      8: (leveldb::DBImpl::BackgroundCompaction()+0x2c0) [0x7f4210]
 646      9: (leveldb::DBImpl::BackgroundCall()+0x68) [0x7f4cc8]
 647      10: ceph-mon() [0x80b3af]
 648      11: (()+0x7e9a) [0x7f4d435d4e9a]
 649      12: (clone()+0x6d) [0x7f4d4196bccd]
 650      NOTE: a copy of the executable, or `objdump -rdS <executable>` is needed to interpret this.
 651
 652     Signed-off-by: Sage Weil <sage@inktank.com>
 653     (cherry picked from commit ea1f316e5de21487ae034a1aa929068ba23ac525)
 654
 655 commit 0143acc49bc5834836afc2c5a9d8f67030bec85f
 656 Author: Sage Weil <sage@inktank.com>
 657 Date:   Tue Jul 2 14:43:17 2013 -0700
 658
 659     sysvinit, upstart: handle symlinks to dirs in /var/lib/ceph/*
 660
 661     Match a symlink to a dir, not just dirs.  This fixes the osd case of e.g.,
 662     creating an osd in /data/osd$id in which ceph-disk makes a symlink from
 663     /var/lib/ceph/osd/ceph-$id.
 664
 665     Fix proposed by Matt Thompson <matt.thompson@mandiant.com>; extended to
 666     include the upstart users too.
 667
 668     Fixes: #5490
 669     Signed-off-by: Sage Weil <sage@inktank.com>
 670     Reviewed-by: Dan Mick <dan.mick@inktank.com>
 671     (cherry picked from commit 87c98e92d1375c8bc76196bbbf06f677bef95e64)
 672
 673 commit 7e878bcc8c1b51538f3c05f854a9dac74c09b116
 674 Author: Sage Weil <sage@inktank.com>
 675 Date:   Mon Jul 1 17:33:11 2013 -0700
 676
 677     rgw: add RGWFormatter_Plain allocation to sidestep cranky strlen()
 678
 679     Valgrind complains about an invalid read when we don't pad the allocation,
 680     and because it is inlined we can't whitelist it for valgrind.  Workaround
 681     the warning by just padding our allocations a bit.
 682
 683     Fixes: #5346
 684     Backport: cuttlefish
 685     Signed-off-by: Sage Weil <sage@inktank.com>
 686     (cherry picked from commit 49ff63b1750789070a8c6fef830c9526ae0f6d9f)
 687
 688 commit ca61402855966210ba1598239eaf454eaad0f5f2
 689 Author: Yan, Zheng <zheng.z.yan@intel.com>
 690 Date:   Wed May 15 11:24:36 2013 +0800
 691
 692     mds: warn on unconnected snap realms
 693
 694     When there are more than one active MDS, restarting MDS triggers
 695     assertion "reconnected_snaprealms.empty()" quite often. If there
 696     is no snapshot in the FS, the items left in reconnected_snaprealms
 697     should be other MDS' mdsdir. I think it's harmless.
 698
 699     If there are snapshots in the FS, the assertion probably can catch
 700     real bugs. But at present, snapshot feature is broken, fixing it is
 701     non-trivial. So replace the assertion with a warning.
 702
 703     Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
 704     (cherry picked from commit 26effc0e583b0a3dade6ec81ef26dec1c94ac8b2)
 705
 706 commit e11f258831e14dc3755e09c0fd4f9bfdf79022a7
 707 Author: Sage Weil <sage@inktank.com>
 708 Date:   Wed Jun 26 06:53:08 2013 -0700
 709
 710     mon/PGMonitor: use post_paxos_update, not init, to refresh from osdmap
 711
 712     We do two things here:
 713      - make init an one-time unconditional init method, which is what the
 714        health service expects/needs.
 715      - switch PGMonitor::init to be post_paxos_update() which is called after
 716        the other services update, which is what PGMonitor really needs.
 717
 718     This is a new version of the fix originally in commit
 719     a2fe0137946541e7b3b537698e1865fbce974ca6 (and those around it).  That is,
 720     this re-fixes a problem where osds do not see pg creates from their
 721     subscribe due to map_pg_creates() not getting called.
 722
 723     Backport: cuttlefish
 724     Signed-off-by: Sage Weil <sage@inktank.com>
 725     (cherry picked from commit e635c47851d185eda557e36bdc4bf3775f7b87a2)
 726
 727     Conflicts:
 728         src/mon/PGMonitor.cc
 729         src/mon/PGMonitor.h
 730
 731 commit 4d07fb014178da3c88edeb8765e1aaacb8cb8ffa
 732 Author: Sage Weil <sage@inktank.com>
 733 Date:   Wed Jun 26 06:52:01 2013 -0700
 734
 735     mon/PaxosService: add post_paxos_update() hook
 736
 737     Some services need to update internal state based on other service's
 738     state, and thus need to be run after everyone has pulled their info out of
 739     paxos.
 740
 741     Backport: cuttlefish
 742     Signed-off-by: Sage Weil <sage@inktank.com>
 743     (cherry picked from commit 131686980f0a930d5de7cbce8234fead5bd438b6)
 744
 745 commit 90f5c448abeb127ae5a5528a79bd7bdbc74cb497
 746 Author: Greg Farnum <greg@inktank.com>
 747 Date:   Thu Jun 27 14:58:14 2013 -0700
 748
 749     ceph-disk: s/else if/elif/
 750
 751     Signed-off-by: Greg Farnum <greg@inktank.com>
 752     Reviewed-by: Joao Luis <joao.luis@inktank.com>
 753     (cherry picked from commit bd8255a750de08c1b8ee5e9c9a0a1b9b16171462)
 754     (cherry picked from commit 9e604ee6943fdb131978afbec51321050faddfc6)
 755
 756 commit 5c4bb463dca5aa61ea5f02f7592d5a3cc82cf6f4
 757 Author: Yehuda Sadeh <yehuda@inktank.com>
 758 Date:   Wed Jun 26 11:28:57 2013 -0700
 759
 760     rgw: fix radosgw-admin buckets list
 761
 762     Fixes: #5455
 763     Backport: cuttlefish
 764     This commit fixes a regression, where radosgw-admin buckets list
 765     operation wasn't returning any data.
 766
 767     Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
 768     Reviewed-by: Sage Weil <sage@inktank.com>
 769     (cherry picked from commit e1f9fe58d2860fcbb18c92d3eb3946236b49a6ce)
 770
 771 commit b2fb48762f32279e73feb83b220339fea31275e9
 772 Author: Sage Weil <sage@inktank.com>
 773 Date:   Wed Jun 19 17:27:49 2013 -0700
 774
 775     ceph-disk: use unix lock instead of lockfile class
 776
 777     The lockfile class relies on file system trickery to get safe mutual
 778     exclusion.  However, the unix syscalls do this for us.  More
 779     importantly, the unix locks go away when the owning process dies, which
 780     is behavior that we want here.
 781
 782     Fixes: #5387
 783     Backport: cuttlefish
 784     Signed-off-by: Sage Weil <sage@inktank.com>
 785     Reviewed-by: Dan Mick <dan.mick@inktank.com>
 786     (cherry picked from commit 2a4953b697a3464862fd3913336edfd7eede2487)
 787
 788 commit 26e7a6fffde4abcb685f34247e8491c05ee2a68d
 789 Author: Sage Weil <sage@inktank.com>
 790 Date:   Wed Jun 26 18:27:49 2013 -0700
 791
 792     ceph-disk: do not mount over an osd directly in /var/lib/ceph/osd/$cluster-$id
 793
 794     If we see a 'ready' file in the target OSD dir, do not mount our device
 795     on top of it.
 796
 797     Among other things, this prevents ceph-disk activate on stray disks from
 798     stepping on teuthology osds.
 799
 800     Fixes: #5445
 801     Signed-off-by: Sage Weil <sage@inktank.com>
 802     (cherry picked from commit 8a17f33b14d858235dfeaa42be1f4842dcfd66d2)
 803
 804 commit ccb3dd5ad5533ca4e9b656b4e3df31025a5f2017
 805 Author: Yan, Zheng <zheng.z.yan@intel.com>
 806 Date:   Tue Apr 2 15:46:51 2013 +0800
 807
 808     mds: fix underwater dentry cleanup
 809
 810     If the underwater dentry is a remove link, we shouldn't mark the
 811     inode clean
 812
 813     Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
 814     (cherry picked from commit 81d073fecb58e2294df12b71351321e6d2e69652)
 815
 816 commit 3020c5ea07a91475a7261dc2b810f5b61a1ae1f2
 817 Author: Sage Weil <sage@inktank.com>
 818 Date:   Mon Jun 24 18:51:07 2013 -0700
 819
 820     mon/Elector: cancel election timer if we bootstrap
 821
 822     If we short-circuit and bootstrap, cancel our timer.  Otherwise it will
 823     go off some time later when we are in who knows what state.
 824
 825     Backport: cuttlefish
 826     Signed-off-by: Sage Weil <sage@inktank.com>
 827     Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
 828     (cherry picked from commit 9ae0ec83dabe37ac15e5165559debdfef7a5f91d)
 829
 830 commit 305f0c50a5f0ffabc73e10bdf4590217d5d5d211
 831 Author: Sage Weil <sage@inktank.com>
 832 Date:   Mon Jun 24 18:12:11 2013 -0700
 833
 834     mon: cancel probe timeout on reset
 835
 836     If we are probing and get (say) an election timeout that calls reset(),
 837     cancel the timer.  Otherwise, we assert later with a splat like
 838
 839     2013-06-24 01:09:33.675882 7fb9627e7700  4 mon.b@0(leader) e1 probe_timeout 0x307a520
 840     2013-06-24 01:09:33.676956 7fb9627e7700 -1 mon/Monitor.cc: In function 'void Monitor::probe_timeout(int)' thread 7fb9627e7700 time 2013-06-24 01:09:43.675904
 841     mon/Monitor.cc: 1888: FAILED assert(is_probing() || is_synchronizing())
 842
 843      ceph version 0.64-613-g134d08a (134d08a9654f66634b893d493e4a92f38acc63cf)
 844      1: (Monitor::probe_timeout(int)+0x161) [0x56f5c1]
 845      2: (Context::complete(int)+0xa) [0x574a2a]
 846      3: (SafeTimer::timer_thread()+0x425) [0x7059a5]
 847      4: (SafeTimerThread::entry()+0xd) [0x7065dd]
 848      5: (()+0x7e9a) [0x7fb966f62e9a]
 849      6: (clone()+0x6d) [0x7fb9652f9ccd]
 850      NOTE: a copy of the executable, or `objdump -rdS <executable>` is needed to interpret this.
 851
 852     Fixes: #5438
 853     Backport: cuttlefish
 854     Signed-off-by: Sage Weil <sage@inktank.com>
 855     Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
 856     (cherry picked from commit 03d3be3eaa96a8e72754c36abd6f355c68d52d59)
 857
 858 commit a8f601d543168f4cdbddf674479d8de4b8dfc732
 859 Author: Alexandre Maragone <alexandre.marangone@inktank.com>
 860 Date:   Tue Jun 18 16:18:01 2013 -0700
 861
 862     ceph-disk: make list_partition behave with unusual device names
 863
 864     When you get device names like sdaa you do not want to mistakenly conclude that
 865     sdaa is a partition of sda.  Use /sys/block/$device/$partition existence
 866     instead.
 867
 868     Fixes: #5211
 869     Backport: cuttlefish
 870     Signed-off-by: Alexandre Maragone <alexandre.maragone@inktank.com>
 871     Reviewed-by: Sage Weil <sage@inktank.com>
 872     (cherry picked from commit 8c0daafe003935881c5192e0b6b59b949269e5ae)
 873
 874 commit 1c890f5cdfc596588e54fffeb016b4a5e9e2124c
 875 Author: Sage Weil <sage@inktank.com>
 876 Date:   Mon Jun 17 20:28:24 2013 -0700
 877
 878     client: fix warning
 879
 880     client/Client.cc: In member function 'virtual void Client::ms_handle_remote_reset(Connection*)':
 881     warning: client/Client.cc:7892:9: enumeration value 'STATE_NEW' not handled in switch [-Wswitch]
 882     warning: client/Client.cc:7892:9: enumeration value 'STATE_OPEN' not handled in switch [-Wswitch]
 883     warning: client/Client.cc:7892:9: enumeration value 'STATE_CLOSED' not handled in switch [-Wswitch]
 884
 885     Signed-off-by: Sage Weil <sage@inktank.com>
 886     Reviewed-by: David Zafman <david.zafman@inktank.com>
 887     (cherry picked from commit 8bd936f077530dfeb2e699164e4492b1c0973088)
 888
 889 commit c3b97591fd8206825bcfe65bdb24fbc75a2a9b42
 890 Author: Sage Weil <sage@inktank.com>
 891 Date:   Mon Jun 24 17:58:48 2013 -0700
 892
 893     mon/AuthMonitor: ensure initial rotating keys get encoded when create_initial called 2x
 894
 895     The create_initial() method may get called multiple times; make sure it
 896     will unconditionally generate new/initial rotating keys.  Move the block
 897     up so that we can easily assert as much.
 898
 899     Broken by commit cd98eb0c651d9ee62e19c2cc92eadae9bed678cd.
 900
 901     Signed-off-by: Sage Weil <sage@inktank.com>
 902     Reviewed-by: Yehuda Sadeh <yehuda@inktank.com>
 903     (cherry picked from commit 521fdc2a4e65559b3da83283e6ca607b6e55406f)
 904
 905 commit 0cc826c385edb2e327505696491d3ff1c3bfe8fd
 906 Author: Sage Weil <sage@inktank.com>
 907 Date:   Mon Jun 24 17:42:04 2013 -0700
 908
 909     init-radosgw.sysv: remove -x debug mode
 910
 911     Fixes: #5443
 912     Signed-off-by: Sage Weil <sage@inktank.com>
 913     (cherry picked from commit 31d6062076fdbcd2691c07a23b381b26abc59f65)
 914
 915 commit 4d57c12faceb7f591f10776c6850d98da55c667b
 916 Author: Sage Weil <sage@inktank.com>
 917 Date:   Mon Jun 24 12:52:44 2013 -0700
 918
 919     common/pick_addresses: behave even after internal_safe_to_start_threads
 920
 921     ceph-mon recently started using Preforker to working around forking issues.
 922     As a result, internal_safe_to_start_threads got set sooner and calls to
 923     pick_addresses() which try to set string config values now fail because
 924     there are no config observers for them.
 925
 926     Work around this by observing the change while we adjust the value.  We
 927     assume pick_addresses() callers are smart enough to realize that their
 928     result will be reflected by cct->_conf and not magically handled elsewhere.
 929
 930     Fixes: #5195, #5205
 931     Backport: cuttlefish
 932     Signed-off-by: Sage Weil <sage@inktank.com>
 933     Reviewed-by: Dan Mick <dan.mick@inktank.com>
 934     (cherry picked from commit eb86eebe1ba42f04b46f7c3e3419b83eb6fe7f9a)
 935
 936 commit e1ac7c6c3ca673d08710829aa5a3c03735710486
 937 Author: Sage Weil <sage@inktank.com>
 938 Date:   Thu Jun 20 15:39:23 2013 -0700
 939
 940     mon/PaxosService: allow paxos service writes while paxos is updating
 941
 942     In commit f985de28f86675e974ac7842a49922a35fe24c6c I mistakenly made
 943     is_writeable() false while paxos was updating due to a misread of
 944     Paxos::propose_new_value() (I didn't see that it would queue).
 945     This is problematic because it narrows the window during which each service
 946     is writeable for no reason.
 947
 948     Allow service to be writeable both when paxos is active and updating.
 949
 950     Signed-off-by: Sage Weil <sage@inktank.com>
 951     (cherry picked from commit 11169693d086e67dcf168ce65ef6e13eebd1a1ab)
 952
 953 commit 02b0b4a9acb439b2ee5deadc8b02492006492931
 954 Author: Sage Weil <sage@inktank.com>
 955 Date:   Fri Jun 7 11:41:21 2013 -0700
 956
 957     mon/PaxosService: not active during paxos UPDATING_PREVIOUS
 958
 959     Treat this as an extension of the recovery process, e.g.
 960
 961      RECOVERING -> ACTIVE
 962     or
 963      RECOVERING -> UPDATING_PREVIOUS -> ACTIVE
 964
 965     and we are not active until we get to "the end" in both cases.
 966
 967     Signed-off-by: Sage Weil <sage@inktank.com>
 968     (cherry picked from commit 392a8e21f8571b410c85be2129ef62dd6fc52b54)
 969
 970 commit c6d5dc4d47838c8c8f4d059b7d018dea3f9c4425
 971 Author: Sage Weil <sage@inktank.com>
 972 Date:   Fri Jun 7 11:40:22 2013 -0700
 973
 974     mon: simplify states
 975
 976     - make states mutually exclusive (an enum)
 977     - rename locked -> updating_previous
 978     - set state prior to begin() to simplify things a bit
 979
 980     Signed-off-by: Sage Weil <sage@inktank.com>
 981     (cherry picked from commit ee34a219605d1943740fdae0d84cfb9020302dd6)
 982
 983 commit c43b1f4dff254df96144b0b4d569cc72421a8fff
 984 Author: Sage Weil <sage@inktank.com>
 985 Date:   Fri Jun 7 11:14:58 2013 -0700
 986
 987     mon/Paxos: not readable when LOCKED
 988
 989     If we are re-proposing a previously accepted value from a previous quorum,
 990     we should not consider it readable, because it is possible it was exposed
 991     to clients as committed (2/3 accepted) but not recored to be committed, and
 992     we do not want to expose old state as readable when new state was
 993     previously readable.
 994
 995     Signed-off-by: Sage Weil <sage@inktank.com>
 996     (cherry picked from commit ec2ea86ed55e00265c2cc5ad0c94460b4c92865c)
 997
 998 commit 10d41200622d76dbf276602828584e7153cb22b5
 999 Author: Sage Weil <sage@inktank.com>
1000 Date:   Fri Jun 7 11:07:38 2013 -0700
1001
1002     mon/Paxos: cleanup: drop unused PREPARING state bit
1003
1004     This is never set when we block, and nobody looks at it.
1005
1006     Signed-off-by: Sage Weil <sage@inktank.com>
1007     (cherry picked from commit 7b7ea8e30e20704caad9a841332ecb2e39819a41)
1008
1009 commit 9d7c40e3f4ea2dd969aa0264ea8a6ad74f3e678a
1010 Author: Sage Weil <sage@inktank.com>
1011 Date:   Thu Jun 6 15:20:05 2013 -0700
1012
1013     mon/PaxosService: simplify is_writeable
1014
1015     Recast this in terms of paxos check + our conditions, and make it
1016     match wait_for_writeable().
1017
1018     Signed-off-by: Sage Weil <sage@inktank.com>
1019     (cherry picked from commit f985de28f86675e974ac7842a49922a35fe24c6c)
1020
1021 commit 35745cba8985c5f3238e3c28fd28b194fae043d9
1022 Author: Sage Weil <sage@inktank.com>
1023 Date:   Tue Jun 4 17:03:15 2013 -0700
1024
1025     mon/PaxosService: simplify readable check
1026
1027     Recast this in terms of the paxos check and our additional conditions,
1028     which match wait_for_readable().
1029
1030     Signed-off-by: Sage Weil <sage@inktank.com>
1031     (cherry picked from commit 3aa61a0beb540e48bf61ceded766d6ff52c95eb2)
1032
1033 commit 57c89291a48c319907fb3029746d9f5a4bd9dd61
1034 Author: Sage Weil <sage@inktank.com>
1035 Date:   Fri May 31 16:45:08 2013 -0700
1036
1037     mon: simplify Monitor::init_paxos()
1038
1039     Signed-off-by: Sage Weil <sage@inktank.com>
1040     (cherry picked from commit e832e76a4af04b091c806ad412bcfd0326d75a2d)
1041
1042 commit fd1769cb2d61e8f2c7921a78760e8f12b28258fb
1043 Author: Sage Weil <sage@inktank.com>
1044 Date:   Fri May 31 16:39:37 2013 -0700
1045
1046     mon/Paxos: go active *after* refreshing
1047
1048     The update_from_paxos() methods occasionally like to trigger new activity.
1049     As long as they check is_readable() and is_writeable(), they will defer
1050     until we go active and that activity will happen in the normal callbacks.
1051
1052     This fixes the problem where we active but is_writeable() is still false,
1053     triggered by PGMonitor::check_osd_map().
1054
1055     Signed-off-by: Sage Weil <sage@inktank.com>
1056     (cherry picked from commit e68b1bd36ed285e38a558899f83cf224d3aa60ed)
1057
1058 commit cf75478d027dfd377424988745230d096dae79ac
1059 Author: Sage Weil <sage@inktank.com>
1060 Date:   Fri May 31 15:32:06 2013 -0700
1061
1062     mon: safely signal bootstrap from MonmapMonitor::update_from_paxos()
1063
1064     Signed-off-by: Sage Weil <sage@inktank.com>
1065     (cherry picked from commit dc83430124a5fd37573202a4cc0986c3c03739ef)
1066
1067 commit 6ac58cd9c1f9c80c5f3cbe97e19cfcd8427db46d
1068 Author: Sage Weil <sage@inktank.com>
1069 Date:   Sun Jun 2 16:57:11 2013 -0700
1070
1071     mon/Paxos: do paxos refresh in finish_proposal; and refactor
1072
1073     Do the paxos refresh inside finish_proposal, ordered *after* the leader
1074     assertion so that MonmapMonitor::update_from_paxos() calling bootstrap()
1075     does not kill us.
1076
1077     Also, remove unnecessary finish_queued_proposal() and move the logic inline
1078     where the bad leader assertion is obvious.
1079
1080     Signed-off-by: Sage Weil <sage@inktank.com>
1081     (cherry picked from commit a42d7582f816b45f5d19c393fd45447555e78fdd)
1082
1083 commit 054e96d96533b1c4078402e43184f13b97329905
1084 Author: Joao Eduardo Luis <joao.luis@inktank.com>
1085 Date:   Sun Jun 2 16:15:02 2013 -0700
1086
1087     mon/PaxosService: cache {first,last}_committed
1088
1089     Refresh the in-memory values when we are told the on-disk paxos state
1090     may have changed.
1091
1092     Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
1093     (cherry picked from commit 2fccb300bdf6ffd44db3462eb05115da11322ed4)
1094
1095 commit 265212a7384399bf85e15e6978bc7543824c0e92
1096 Author: Sage Weil <sage@inktank.com>
1097 Date:   Fri May 31 14:30:48 2013 -0700
1098
1099     mon: no need to refresh from _active
1100
1101     The refresh is done explicitly by the monitor, independent of the more
1102     fragile PaxosService callbacks.
1103
1104     Signed-off-by: Sage Weil <sage@inktank.com>
1105     (cherry picked from commit d941363d6e4249e97b64faff0e573f75e918ac0c)
1106
1107 commit 1d8662504299babec22c714662cefbb86a0acb8b
1108 Author: Sage Weil <sage@inktank.com>
1109 Date:   Sun Jun 2 16:10:57 2013 -0700
1110
1111     mon: remove unnecessary update_from_paxos calls
1112
1113     The refresh() will do this when the state changes; no need to
1114     opportunistically call this method all of the time.
1115
1116     Signed-off-by: Sage Weil <sage@inktank.com>
1117     (cherry picked from commit 03014a4ecc06cde420fad0c6c2a0177ebd7b839d)
1118
1119 commit 34acc5a3161b6bcda2b9f7ce18d89a8618fff1c5
1120 Author: Sage Weil <sage@inktank.com>
1121 Date:   Sun Jun 2 16:14:01 2013 -0700
1122
1123     mon: explicitly refresh_from_paxos() when leveldb state changes
1124
1125     Instead of opportunistically calling each service's update_from_paxos(),
1126     instead explicitly refresh all in-memory state whenever we know the
1127     paxos state may have changed.  This is simpler and less fragile.
1128
1129     Signed-off-by: Sage Weil <sage@inktank.com>
1130     (cherry picked from commit cc339c07312006e65854207523f50542d00ecf87)
1131
1132 commit 4474a0cc6c009a566ecf46efadb39d80343a7c68
1133 Author: Sage Weil <sage@inktank.com>
1134 Date:   Sun Jun 23 09:25:55 2013 -0700
1135
1136     mon/AuthMonitor: make initial auth include rotating keys
1137
1138     This closes a very narrow race during mon creation where there are no
1139     service keys.
1140
1141     Fixes: #5427
1142     Signed-off-by: Sage Weil <sage@inktank.com>
1143     (cherry picked from commit cd98eb0c651d9ee62e19c2cc92eadae9bed678cd)
1144
1145 commit d572cf6f77418f217a5a8e37f1124dc566e24d0b
1146 Author: Sage Weil <sage@inktank.com>
1147 Date:   Fri Jun 21 11:53:29 2013 -0700
1148
1149     mds: fix iterator invalidation for backtrace removal
1150
1151     - Don't increment before we dereference!
1152     - We need to update the iterator before we delete the item.
1153
1154     This code is changed in master, so this fix is for cuttlefish only.
1155
1156     Signed-off-by: Sage Weil <sage@inktank.com>
1157     Reviewed-by: Greg Farnum <greg@inktank.com>
1158
1159 commit 50957772c3582290331f69ba4a985b1cdf86834d
1160 Author: Sage Weil <sage@inktank.com>
1161 Date:   Thu May 9 09:44:20 2013 -0700
1162
1163     osd: init test_ops_hook
1164
1165     CID 1019628 (#1 of 1): Uninitialized pointer field (UNINIT_CTOR)
1166     2. uninit_member: Non-static class member "test_ops_hook" is not initialized in this constructor nor in any functions that it calls.
1167
1168     Signed-off-by: Sage Weil <sage@inktank.com>
1169     (cherry picked from commit e30a03210c3efb768b1653df5ae58917ef26e579)
1170
1171 commit 17d2745f095e7bb640dece611d7824d370ea3b81
1172 Author: Sage Weil <sage@inktank.com>
1173 Date:   Thu May 9 09:45:51 2013 -0700
1174
1175     osd: initialize OSDService::next_notif_id
1176
1177     CID 1019627 (#1 of 1): Uninitialized scalar field (UNINIT_CTOR)
1178     2. uninit_member: Non-static class member "next_notif_id" is not initialized in this constructor nor in any functions that it calls.
1179
1180     Signed-off-by: Sage Weil <sage@inktank.com>
1181     (cherry picked from commit 499edd8bfc355c2d590f5fa1ef197d1ea5680351)
1182
1183 commit ffdb7236a994aa20b5f75860b9c81dac0f131f9a
1184 Author: Sage Weil <sage@inktank.com>
1185 Date:   Thu Jun 20 09:46:42 2013 -0700
1186
1187     mon: more fix dout use in sync_requester_abort()
1188
1189     Signed-off-by: Sage Weil <sage@inktank.com>
1190     (cherry picked from commit d60534b8f59798feaeeaa17adba2a417d7777cbf)
1191
1192 commit 38ddae04bb974a93f1718c509363f1afbe6b612d
1193 Author: Sage Weil <sage@inktank.com>
1194 Date:   Mon Jun 10 11:48:25 2013 -0700
1195
1196     mon: fix raw use of *_dout in sync_requester_abort()
1197
1198     Signed-off-by: Sage Weil <sage@inktank.com>
1199     (cherry picked from commit 8a4ed58e39b287fd8667c62b45848487515bdc80)