ceph/PendingReleaseNotes

   1 >= 12.0.0
   2 ------
   3 * The "journaler allow split entries" config setting has been removed.
   4 * The 'apply' mode of cephfs-journal-tool has been removed
   5
   6 12.0.0
   7 ------
   8
   9  * When assigning a network to the public network and not to
  10    the cluster network the network specification of the public
  11    network will be used for the cluster network as well.
  12    In older versions this would lead to cluster services
  13    being bound to 0.0.0.0:<port>, thus making the
  14    cluster service even more publicly available than the
  15    public services. When only specifying a cluster network it
  16    will still result in the public services binding to 0.0.0.0.
  17
  18 *  Some variants of the omap_get_keys and omap_get_vals librados
  19    functions have been deprecated in favor of omap_get_vals2 and
  20    omap_get_keys2.  The new methods include an output argument
  21    indicating whether there are additional keys left to fetch.
  22    Previously this had to be inferred from the requested key count vs
  23    the number of keys returned, but this breaks with new OSD-side
  24    limits on the number of keys or bytes that can be returned by a
  25    single omap request.  These limits were introduced by kraken but
  26    are effectively disabled by default (by setting a very large limit
  27    of 1 GB) because users of the newly deprecated interface cannot
  28    tell whether they should fetch more keys or not.  In the case of
  29    the standalone calls in the C++ interface
  30    (IoCtx::get_omap_{keys,vals}), librados has been updated to loop on
  31    the client side to provide a correct result via multiple calls to
  32    the OSD.  In the case of the methods used for building
  33    multi-operation transactions, however, client-side looping is not
  34    practical, and the methods have been deprecated.  Note that use of
  35    either the IoCtx methods on older librados versions or the
  36    deprecated methods on any version of librados will lead to
  37    incomplete results if/when the new OSD limits are enabled.
  38
  39 * In previous versions, if a client sent an op to the wrong OSD, the OSD
  40   would reply with ENXIO.  The rationale here is that the client or OSD is
  41   clearly buggy and we want to surface the error as clearly as possible.
  42   We now only send the ENXIO reply if the osd_enxio_on_misdirected_op option
  43   is enabled (it's off by default).  This means that a VM using librbd that
  44   previously would have gotten an EIO and gone read-only will now see a
  45   blocked/hung IO instead.
  46
  47 *  When configuring ceph-fuse mounts in /etc/fstab, a new syntax is
  48    available that uses "ceph.<arg>=<val>" in the options column, instead
  49    of putting configuration in the device column.  The old style syntax
  50    still works.  See the documentation page "Mount CephFS in your
  51    file systems table" for details.
  52
  53 12.0.1
  54 ------
  55
  56 * The original librados rados_objects_list_open (C) and objects_begin
  57   (C++) object listing API, deprecated in Hammer, has finally been
  58   removed.  Users of this interface must update their software to use
  59   either the rados_nobjects_list_open (C) and nobjects_begin (C++) API or
  60   the new rados_object_list_begin (C) and object_list_begin (C++) API
  61   before updating the client-side librados library to Luminous.
  62
  63   Object enumeration (via any API) with the latest librados version
  64   and pre-Hammer OSDs is no longer supported.  Note that no in-tree
  65   Ceph services rely on object enumeration via the deprecated APIs, so
  66   only external librados users might be affected.
  67
  68   The newest (and recommended) rados_object_list_begin (C) and
  69   object_list_begin (C++) API is only usable on clusters with the
  70   SORTBITWISE flag enabled (Jewel and later).  (Note that this flag is
  71   required to be set before upgrading beyond Jewel.)
  72
  73 * The rados copy-get-classic operation has been removed since it has not been
  74   used by the OSD since before hammer.  It is unlikely any librados user is
  75   using this operation explicitly since there is also the more modern copy-get.
  76
  77 * The RGW api for getting object torrent has changed its params from 'get_torrent'
  78   to 'torrent' so that it can be compatible with Amazon S3. Now the request for
  79   object torrent is like 'GET /ObjectName?torrent'.
  80
  81 * The configuration option "osd pool erasure code stripe width" has
  82   been replaced by "osd pool erasure code stripe unit", and given the
  83   ability to be overridden by the erasure code profile setting
  84   "stripe_unit". For more details see "Erasure Code Profiles" in the
  85   documentation.
  86
  87 * rbd and cephfs can use erasure coding with bluestore. This may be
  88   enabled by setting 'allow_ec_overwrites' to 'true' for a pool. Since
  89   this relies on bluestore's checksumming to do deep scrubbing,
  90   enabling this on a pool stored on filestore is not allowed.
  91
  92 * The 'rados df' JSON output now prints numeric values as numbers instead of
  93   strings.
  94
  95 * There was a bug introduced in Jewel (#19119) that broke the mapping behavior
  96   when an "out" OSD that still existed in the CRUSH map was removed with 'osd rm'.
  97   This could result in 'misdirected op' and other errors.  The bug is now fixed,
  98   but the fix itself introduces the same risk because the behavior may vary between
  99   clients and OSDs.  To avoid problems, please ensure that all OSDs are removed
 100   from the CRUSH map before deleting them.  That is, be sure to do::
 101
 102      ceph osd crush rm osd.123
 103
 104   before::
 105
 106      ceph osd rm osd.123
 107
 108 12.0.2
 109 ------
 110
 111 * The original librados rados_objects_list_open (C) and objects_begin
 112   (C++) object listing API, deprecated in Hammer, has finally been
 113   removed.  Users of this interface must update their software to use
 114   either the rados_nobjects_list_open (C) and nobjects_begin (C++) API or
 115   the new rados_object_list_begin (C) and object_list_begin (C++) API
 116   before updating the client-side librados library to Luminous.
 117
 118   Object enumeration (via any API) with the latest librados version
 119   and pre-Hammer OSDs is no longer supported.  Note that no in-tree
 120   Ceph services rely on object enumeration via the deprecated APIs, so
 121   only external librados users might be affected.
 122
 123   The newest (and recommended) rados_object_list_begin (C) and
 124   object_list_begin (C++) API is only usable on clusters with the
 125   SORTBITWISE flag enabled (Jewel and later).  (Note that this flag is
 126   required to be set before upgrading beyond Jewel.)
 127 * CephFS clients without the 'p' flag in their authentication capability
 128   string will no longer be able to set quotas or any layout fields.  This
 129   flag previously only restricted modification of the pool and namespace
 130   fields in layouts.
 131 * CephFS directory fragmentation (large directory support) is enabled
 132   by default on new filesystems.  To enable it on existing filesystems
 133   use "ceph fs set <fs_name> allow_dirfrags".
 134 * CephFS will generate a health warning if you have fewer standby daemons
 135   than it thinks you wanted.  By default this will be 1 if you ever had
 136   a standby, and 0 if you did not.  You can customize this using
 137   ``ceph fs set <fs> standby_count_wanted <number>``.  Setting it
 138   to zero will effectively disable the health check.
 139 * The "ceph mds tell ..." command has been removed.  It is superceded
 140   by "ceph tell mds.<id> ..."
 141
 142 12.1.0
 143 ------
 144
 145 * The ``mon_osd_max_op_age`` option has been renamed to
 146   ``mon_osd_warn_op_age`` (default: 32 seconds), to indicate we
 147   generate a warning at this age.  There is also a new
 148   ``mon_osd_err_op_age_ratio`` that is a expressed as a multitple of
 149   ``mon_osd_warn_op_age`` (default: 128, for roughly 60 minutes) to
 150   control when an error is generated.
 151
 152 * The default maximum size for a single RADOS object has been reduced from
 153   100GB to 128MB.  The 100GB limit was completely impractical in practice
 154   while the 128MB limit is a bit high but not unreasonable.  If you have an
 155   application written directly to librados that is using objects larger than
 156   128MB you may need to adjust ``osd_max_object_size``.
 157
 158 * The semantics of the 'rados ls' and librados object listing
 159   operations have always been a bit confusing in that "whiteout"
 160   objects (which logically don't exist and will return ENOENT if you
 161   try to access them) are included in the results.  Previously
 162   whiteouts only occurred in cache tier pools.  In luminous, logically
 163   deleted but snapshotted objects now result in a whiteout object, and
 164   as a result they will appear in 'rados ls' results, even though
 165   trying to read such an object will result in ENOENT.  The 'rados
 166   listsnaps' operation can be used in such a case to enumerate which
 167   snapshots are present.
 168
 169   This may seem a bit strange, but is less strange than having a
 170   deleted-but-snapshotted object not appear at all and be completely
 171   hidden from librados's ability to enumerate objects.  Future
 172   versions of Ceph will likely include an alternative object
 173   enumeration interface that makes it more natural and efficient to
 174   enumerate all objects along with their snapshot and clone metadata.
 175
 176 * The deprecated 'crush_ruleset' property has finally been removed; please use
 177   'crush_rule' instead for the 'osd pool get ...' and 'osd pool set ..' commands.
 178
 179 * The 'osd pool default crush replicated ruleset' option has been
 180   removed and replaced by the 'osd pool default crush rule' option.
 181   By default it is -1, which means the mon will pick the first type
 182   replicated rule in the CRUSH map for replicated pools.  Erasure
 183   coded pools have rules that are automatically created for them if they are
 184   not specified at pool creation time.