ceph/doc/changelog/v0.61.7.txt

   1 commit 8f010aff684e820ecc837c25ac77c7a05d7191ff
   2 Author: Gary Lowell <gary.lowell@inktank.com>
   3 Date:   Wed Jul 24 20:44:12 2013 -0700
   4
   5     v0.61.7
   6
   7 commit 24a56a9637afd8c64b71d264359c78a25d52be02
   8 Author: Sage Weil <sage@inktank.com>
   9 Date:   Wed Jul 24 14:46:24 2013 -0700
  10
  11     ceph-disk: use new get_dev_path helper for list
  12
  13     Backport: cuttlefish
  14     Signed-off-by: Sage Weil <sage@inktank.com>
  15     Reviewed-by: Dan Mick <dan.mick@inktank.com>
  16     Tested-by: Olivier Bonvalet <ob.ceph@daevel.fr>
  17     (cherry picked from commit fd1fd664d6102a2a96b27e8ca9933b54ac626ecb)
  18
  19 commit 1f8e4b15eeb132fd7f389318009b19f8f13adbf5
  20 Author: Sage Weil <sage@inktank.com>
  21 Date:   Thu Jul 11 12:59:56 2013 -0700
  22
  23     ceph-disk: use /sys/block to determine partition device names
  24
  25     Not all devices are basename + number; some have intervening character(s),
  26     like /dev/cciss/c0d1p2.
  27
  28     Signed-off-by: Sage Weil <sage@inktank.com>
  29     (cherry picked from commit 2ea8fac441141d64ee0d26c5dd2b441f9782d840)
  30
  31 commit 0a08c8226cb3e461301beade9bab2e264d1b960e
  32 Author: Sage Weil <sage@inktank.com>
  33 Date:   Wed Jul 3 11:01:58 2013 -0700
  34
  35     ceph-disk: reimplement is_partition() using /sys/block
  36
  37     Signed-off-by: Sage Weil <sage@inktank.com>
  38     (cherry picked from commit 5b031e100b40f597752b4917cdbeebb366eb98d7)
  39
  40 commit 056000346db09ea7274a22e57cf4b86a7ea4090e
  41 Author: Sage Weil <sage@inktank.com>
  42 Date:   Wed Jul 3 11:01:39 2013 -0700
  43
  44     ceph-disk: use get_dev_name() helper throughout
  45
  46     This is more robust than the broken split trick.
  47
  48     Signed-off-by: Sage Weil <sage@inktank.com>
  49     (cherry picked from commit 3359aaedde838c98d1155611e157fd2da9e8b9f5)
  50
  51 commit f3ee3e690c42769229a6cd9ae8dec43f2aa22ecd
  52 Author: Sage Weil <sage@inktank.com>
  53 Date:   Wed Jul 3 10:55:36 2013 -0700
  54
  55     ceph-disk: refactor list_[all_]partitions
  56
  57     Make these methods work in terms of device *names*, not paths, and fix up
  58     the only direct list_partitions() caller to do the same.
  59
  60     Signed-off-by: Sage Weil <sage@inktank.com>
  61     (cherry picked from commit 35d3f2d84808efda3d2ac868afe03e6959d51c03)
  62
  63 commit be12811b4cb98ff1c2c691c67af7ad3586c436ff
  64 Author: Sage Weil <sage@inktank.com>
  65 Date:   Wed Jul 3 10:52:29 2013 -0700
  66
  67     ceph-disk: add get_dev_name, path helpers
  68
  69     Signed-off-by: Sage Weil <sage@inktank.com>
  70     (cherry picked from commit e0401591e352ea9653e3276d66aebeb41801eeb3)
  71
  72 commit f46dbc462f623e9ab6c00394abb4d890e5d90890
  73 Author: Sage Weil <sage@inktank.com>
  74 Date:   Tue Jun 18 16:21:48 2013 -0700
  75
  76     ceph-disk: handle /dev/foo/bar devices throughout
  77
  78     Assume the last component is the unique device name, even if it appears
  79     under a subdir of /dev.
  80
  81     Signed-off-by: Sage Weil <sage@inktank.com>
  82     (cherry picked from commit cb97338b1186939deecb78e9d949c38c3ef59026)
  83
  84 commit f799dac7bdf7cf0824a177131473cf59ef3c5205
  85 Author: Sage Weil <sage@inktank.com>
  86 Date:   Mon Jun 17 20:54:15 2013 -0700
  87
  88     ceph-disk: make is_held() smarter about full disks
  89
  90     Handle the case where the device is a full disk.  Make the partition
  91     check a bit more robust (don't make assumptions about naming aside from
  92     the device being a prefix of the partition).
  93
  94     Signed-off-by: Sage Weil <sage@inktank.com>
  95     (cherry picked from commit e082f1247fb6ddfb36c4223cbfdf500d6b45c978)
  96
  97 commit 27f31895664fa7f10c1617d486f2a6ece0f97091
  98 Author: Sage Weil <sage@inktank.com>
  99 Date:   Wed Jul 24 11:55:42 2013 -0700
 100
 101     mon/OSDMonitor: search for latest full osdmap if record version is missing
 102
 103     In 97462a3213e5e15812c79afc0f54d697b6c498b1 we tried to search for a
 104     recent full osdmap but were looking at the wrong key.  If full_0 was
 105     present we could record that the latest full map was last_committed even
 106     though it wasn't present.  This is fixed in 76cd7ac1c, but we need to
 107     compensate for when get_version_latest_full() gives us a back version
 108     number by repeating the search.
 109
 110     Fixes: #5737
 111     Signed-off-by: Sage Weil <sage@inktank.com>
 112     Reviewed-by: Joao Eduardo Luis <joao.luis@inktank.com>
 113     (cherry picked from commit c2131d4047156aa2964581c9dbd93846382a07e7)
 114
 115 commit 5b0967f03efb1be210b52f24f095f023fe1bc539
 116 Author: Joao Eduardo Luis <joao.luis@inktank.com>
 117 Date:   Mon Jun 17 14:43:36 2013 +0100
 118
 119     test: test_store_tool: global init before using LevelDBStore
 120
 121     Fixes a segfault
 122
 123     Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
 124     Reviewed-by: Sage Weil <sage@inktank.com>
 125     (cherry picked from commit a7a7d3fc8a2ba4a30ef136a32f2903d157b3e19a)
 126
 127 commit 115468c73f121653eec2efc030d5ba998d834e43
 128 Author: Joao Eduardo Luis <joao.luis@inktank.com>
 129 Date:   Wed Jul 24 12:00:28 2013 +0100
 130
 131     mon: OSDMonitor: fix a bug introduced on 97462a32
 132
 133     Fixes: #5737
 134     Backport: cuttlefish
 135
 136     Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
 137     Reviewed-by: Sage Weil <sage@inktank.com>
 138     (cherry picked from commit 76cd7ac1c2094b34ad36bea89b2246fa90eb2f6d)
 139
 140 commit 938a639e2cb6abd22c2c588e619c1aae32c6521f
 141 Author: Sage Weil <sage@inktank.com>
 142 Date:   Sun Jul 21 08:48:18 2013 -0700
 143
 144     mon/Paxos: fix pn for uncommitted value during collect/last phase
 145
 146     During the collect/last exchange, peers share any uncommitted values
 147     with the leader.  They are supposed to also share the pn under which
 148     that value was accepted, but were instead using the just-accepted pn
 149     value.  This effectively meant that we *always* took the uncommitted
 150     value; if there were multiples, which one we accepted depended on what
 151     order the LAST messages arrived, not which pn the values were generated
 152     under.
 153
 154     The specific failure sequence I observed:
 155
 156      - collect
 157       - learned uncommitted value for 262 from myself
 158       - send collect with pn 901
 159      - got last with pn 901 (incorrect) for 200 (old) from peer
 160       - discard our own value, remember the other
 161      - finish collect phase
 162       - ignore old uncommitted value
 163
 164     Fix this by storing a pending_v and pending_pn value whenever we accept
 165     a value.  Use this to send an appropriate pn value in the LAST reply
 166     so that the leader can make it's decision about which uncommitted value
 167     to accept based on accurate information.  Also use it when we learn
 168     the uncommitted value from ourselves.
 169
 170     We could probably be more clever about storing less information here,
 171     for example by omitting pending_v and clearing pending_pn at the
 172     appropriate point, but that would be more fragile.  Similarly, we could
 173     store a pn for *every* commit if we wanted to lay some groundwork for
 174     having multiple uncommitted proposals in flight, but I don't want to
 175     speculate about what is necessary or sufficient for a correct solution
 176     there.
 177
 178     Fixes: #5698
 179     Backport: cuttlefish, bobtail
 180     Signed-off-by: Sage Weil <sage@inktank.com>
 181     (cherry picked from commit 20baf662112dd5f560bc3a2d2114b469444c3de8)
 182
 183 commit 18596340f020be1f21bdc9bcc752ae1da4a93a46
 184 Author: Sage Weil <sage@inktank.com>
 185 Date:   Sun Jul 21 08:12:46 2013 -0700
 186
 187     mon/Paxos: debug ignored uncommitted values
 188
 189     Signed-off-by: Sage Weil <sage@inktank.com>
 190     (cherry picked from commit 19b29788966eb80ed847630090a16a3d1b810969)
 191
 192 commit f598245f1355d7791162c03d90bdd97b013e56f3
 193 Author: Sage Weil <sage@inktank.com>
 194 Date:   Sun Jul 21 08:11:22 2013 -0700
 195
 196     mon/Paxos: only learn uncommitted value if it is in the future
 197
 198     If an older peer sends an uncommitted value, make sure we only take it
 199     if it is in the future, and at least as new as any current uncommitted
 200     value.
 201
 202     (Prior to the previous patch, peers could send values from long-past
 203     rounds.  The pn values are also bogus.)
 204
 205     Signed-off-by: Sage Weil <sage@inktank.com>
 206     (cherry picked from commit b3253a453c057914753846c77499f98d3845c58e)
 207
 208 commit 732286a28cd8a643593d490a7a84a590d372f78d
 209 Author: Sage Weil <sage@inktank.com>
 210 Date:   Mon Jul 22 14:13:23 2013 -0700
 211
 212     mon/Paxos: only share uncommitted value if it is next
 213
 214     We may have an uncommitted value from our perspective (it is our lc + 1)
 215     when the collector has a much larger lc (because we have been out for
 216     the last few rounds).  Only share an uncommitted value if it is in fact
 217     the next value.
 218
 219     Signed-off-by: Sage Weil <sage@inktank.com>
 220     (cherry picked from commit b26b7f6e5e02ac6beb66e3e34e177e6448cf91cf)