ceph/doc/changelog/v0.72.2.txt

   1 commit a913ded2ff138aefb8cb84d347d72164099cfd60
   2 Author: Gary Lowell <gary.lowell@inktank.com>
   3 Date:   Fri Dec 20 19:28:37 2013 +0000
   4
   5     v0.72.2
   6
   7 commit b570c8014325fef590d16e6157cb33026155932c
   8 Author: Yehuda Sadeh <yehuda@inktank.com>
   9 Date:   Wed Dec 18 13:11:01 2013 -0800
  10
  11     rgw: fix use-after-free when releasing completion handle
  12
  13     Backport: emperor, dumpling
  14     Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
  15     Reviewed-by: Sage Weil <sage@inktank.com>
  16     (cherry picked from commit c8890ab2d46fe8e12200a0d2f9eab31c461fb871)
  17
  18 commit 451381b4dba9d50fbee2accf5d75562615001ef3
  19 Author: Yehuda Sadeh <yehuda@inktank.com>
  20 Date:   Wed Dec 18 13:10:21 2013 -0800
  21
  22     rgw: don't return data within the librados cb
  23
  24     Fixes: #7030
  25     The callback is running within a single Finisher thread, thus we
  26     shouldn't block there. Append read data to a list and flush it within
  27     the iterate context.
  28
  29     Reviewed-by: Sage Weil <sage@inktank.com>
  30     Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
  31     (cherry picked from commit d6a4f6adfaa75c3140d07d6df7be03586cc16183)
  32
  33 commit c044b218cb32d7ef2f2aa5db81ebf38ad0dc1049
  34 Author: Sage Weil <sage@inktank.com>
  35 Date:   Sun Dec 1 22:21:31 2013 -0800
  36
  37     Partial revert "mon: osd pool set syntax relaxed, modify unit tests"
  38
  39     This reverts commit 08327fed8213a5d24cd642e12b38a171b98924cb, except
  40     for the hashpspool bit.  We switched back to an integer argument in
  41     commit 337195f04653eed8e8f153a5b074f3bd48408998.
  42
  43     Signed-off-by: Sage Weil <sage@inktank.com>
  44     (cherry picked from commit e80ab94bf44e102fcd87d16dc11e38ca4c0eeadb)
  45     Reviewed-by: Greg Farnum <greg@inktank.com>
  46
  47 commit 1983ed8013637e9b80ca4a0e8d2800d6a36b120e
  48 Author: Joao Eduardo Luis <joao.luis@inktank.com>
  49 Date:   Tue Nov 19 23:21:11 2013 +0000
  50
  51     mon: OSDMonitor: drop cmdval_get() for unused variable
  52
  53     We don't ever use any value as a float, so just drop obtaining it.  This
  54     makes it easier to partially revert 2fe0d0d9 in an upcoming patch.
  55
  56     Backport: emperor
  57
  58     Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
  59     (cherry picked from commit 7191bb2b2485c7819ca7b9d9434d803d0c94db7a)
  60     Reviewed-by: Sage Weil <sage@inktank.com>
  61     Reviewed-by: Greg Farnum <greg@inktank.com>
  62
  63 commit 1c4759cebb11ae1aa84aa3cbdb2b31e6d111f28e
  64 Author: Joao Eduardo Luis <joao.luis@inktank.com>
  65 Date:   Fri Nov 22 02:10:35 2013 +0000
  66
  67     mon: OSDMonitor: receive CephInt on 'osd pool set' instead on CephString
  68
  69     This partially reverts 2fe0d0d9 in order to allow Emperor monitors to
  70     forward mon command messages to Dumpling monitors without breaking a
  71     cluster.
  72
  73     The need for this patch became obvious after issue #6796 was triggered.
  74     Basically, in a mixed cluster of Emperor/Dumpling monitors, if a client
  75     happens to obtain the command descriptions from an Emperor monitor and
  76     then issue an 'osd pool set' this can turn out in one of two ways:
  77
  78     1. client msg gets forwarded to an Emperor leader and everything's a-okay;
  79     2. client msg gets forwarded to a Dumpling leader and the string fails to
  80     be interpreted without the monitor noticing, thus leaving the monitor with
  81     an uninitialized variable leading to trouble.
  82
  83     If 2 is triggered, a multitude of bad things can happen, such as thousands
  84     of pg splits, due to a simple 'osd set pool foo pg_num 128' turning out
  85     to be interpreted as 109120394 or some other random number.
  86
  87     This patch is such that we make sure the client sends an integer instead
  88     of a string. We also make sure to interpret anything the client sends as
  89     possibly being a string, or an integer.
  90
  91     Fixes: 6796
  92     Backport: emperor
  93
  94     Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
  95     (cherry picked from commit 337195f04653eed8e8f153a5b074f3bd48408998)
  96     Reviewed-by: Sage Weil <sage@inktank.com>
  97     Reviewed-by: Greg Farnum <greg@inktank.com>
  98
  99 commit c2e467b42dfaecf082a3838b10e873f9f8bd5633
 100 Author: Josh Durgin <josh.durgin@inktank.com>
 101 Date:   Mon Nov 25 13:43:43 2013 -0800
 102
 103     init, upstart: prevent daemons being started by both
 104
 105     There can be only one init system starting a daemon. If there is a
 106     host entry in ceph.conf for a daemon, sysvinit would try to start it
 107     even if the daemon's directory did not include a sysvinit file. This
 108     preserves backwards compatibility with older installs using sysvinit,
 109     but if an upstart file is present in the daemon's directory, upstart
 110     will try to start them, regardless of host entries in ceph.conf.
 111
 112     If there's an upstart file in a daemon's directory and a host entry
 113     for that daemon in ceph.conf, both sysvinit and upstart would attempt
 114     to manage it.
 115
 116     Fix this by only starting daemons if the marker file for the other
 117     init system is not present. This maintains backwards compatibility
 118     with older installs using neither sysvinit or upstart marker files,
 119     and does not break any valid configurations. The only configuration
 120     that would break is one with both sysvinit and upstart files present
 121     for the same daemon.
 122
 123     Backport: emperor, dumpling
 124     Reported-by: Tim Spriggs <tims@uahirise.org>
 125     Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
 126     (cherry picked from commit 5e34beb61b3f5a1ed4afd8ee2fe976de40f95ace)
 127
 128 commit 8812c4e958b154ca401c7257fd48ccaffe013639
 129 Author: Yehuda Sadeh <yehuda@inktank.com>
 130 Date:   Wed Nov 27 13:34:00 2013 -0800
 131
 132     rgw: don't error out on empty owner when setting acls
 133
 134     Fixes: #6892
 135     Backport: dumpling, emperor
 136     s3cmd specifies empty owner field when trying to set acls on object
 137     / bucket. We errored out as it didn't match the current owner name, but
 138     with this change we ignore it.
 139
 140     Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
 141     (cherry picked from commit 14cf4caff58cc2c535101d48c53afd54d8632104)
 142
 143 commit 69e055f367ae37dd771c050f158d684e2a3cb246
 144 Author: Yehuda Sadeh <yehuda@inktank.com>
 145 Date:   Fri Nov 22 07:04:01 2013 -0800
 146
 147     rgw: lower some debug message
 148
 149     Fixes: #6084
 150     Backport: dumpling, emperor
 151
 152     Reported-by: Ron Allred <rallred@itrefined.com>
 153     Signed-off-by: Yehuda Sadeh <yehuda@inktank.com>
 154     (cherry picked from commit b35fc1bc2ec8c8376ec173eb1c3e538e02c1694e)
 155
 156 commit 53951fd5459d6396b6ab59b52dc33df0ebcf4414
 157 Author: Samuel Just <sam.just@inktank.com>
 158 Date:   Tue Nov 12 15:15:26 2013 -0800
 159
 160     ReplicatedPG: test for missing head before find_object_context
 161
 162     find_object_context doesn't return EAGAIN for a missing head.
 163     I chose not to change that behavior since it might hide bugs
 164     in the future.  All other callers check for missing on head
 165     before calling into find_object_context because we potentially
 166     need head or snapdir to map a snapid onto a clone.
 167
 168     Backport: emperor
 169     Fixes: 6758
 170     Signed-off-by: Samuel Just <sam.just@inktank.com>
 171     Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
 172     Reviewed-by: David Zafman <david.zafman@inktank.com>
 173     (cherry picked from commit dd9d8b020286d5e3a69455023c3724a7b436d687)
 174
 175 commit 83ee6843de6432278dc2891526691006c51eb1fa
 176 Author: Josh Durgin <josh.durgin@inktank.com>
 177 Date:   Mon Nov 18 14:39:12 2013 -0800
 178
 179     osd: fix bench block size
 180
 181     The command was declared to take 'size' in dumpling, but was trying to
 182     read 'bsize' instead, so it always used the default of 4MiB. Change
 183     the bench command to read 'size', so it matches what existing clients
 184     are sending.
 185
 186     Fixes: #6795
 187     Backport: emperor, dumpling
 188     Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
 189     (cherry picked from commit 40a76ef0d09f8ecbea13712410d9d34f25b91935)