aainscow · May 5, 2026 · Jul 9, 2026 · Oct 29, 2026 · Oct 13, 2026 · Oct 13, 2026
@@ -42,6 +42,41 @@
   be removed in the X release. People wanting to continue using
   Elasticsearch can populate it with bucket notifications. The
   upcoming S3 Inventory feature also covers some of the use case.
+
+* RADOS: Erasure-coded (FastEC) and replicated pools can be stretched over two
+  zones, one pool at a time, with ``ceph osd pool create --num_zones 2`` or
+  ``ceph osd pool set <pool> num_zones 2``. See "Upgrading from Earlier
+  Releases" in :ref:`stretch_mode`.
+
+  - When ``require_osd_release`` is raised to this release, each pool of a
+    cluster stretched with ``ceph mon enable_stretch_mode`` gets two zones and
+    a ``min_size`` counted in each zone (2 becomes 1, 3 becomes 2).
+  - From then on, ``osd_pool_default_size`` counts the copies in each zone of
+    a stretched pool, replacing ``mon_stretch_pool_size``. Only the value in
+    the ``global`` section of the configuration database is converted. A value
+    set in a Monitor's ``ceph.conf`` or in a ``mon`` section is not, and gives
+    new pools that many copies in each zone. The new
+    ``POOL_DEFAULT_CONFIG_OVERRIDE`` health warning reports such a value:
+    remove it.
+
+* RADOS: ``ceph osd pool create`` takes named parameters, such as
+  ``--pool_type``, ``--replica``, ``--k`` and ``--m``, and still accepts the
+  positional ones. A parameter that is not given comes from a pool creation
+  default, which ``ceph osd pool default set`` and ``ceph osd pool default
+  get`` manage. See :ref:`pool-creation-defaults`.
+
+* RADOS: An erasure code profile or CRUSH rule, other than the defaults, is
+  now deleted with the last pool that uses it. Scripts that reuse one after
+  deleting its pools must create it again.
+
+* RADOS: Changing an erasure code profile that a pool uses now needs
+  ``--yes-i-really-mean-it``: ``--force`` is no longer enough.
+
+* RADOS: When ``require_osd_release`` is raised to this release, each
+  erasure-coded pool that uses the ``default`` erasure code profile gets a
+  profile of its own, such as ``mypool-k2-m2``, as does each new pool created
+  without a named profile.
+
 * RGW: The default value of ``rgw_rest_conn_connect_to_resolved_ips`` changed
   from ``false`` to ``true``.
 

@@ -67,6 +67,7 @@ Synopsis
         [--pool-snaps]
         [--balance-reads]
         [--localize-reads]
+        [--crush-location <loc>]
         [--offlen_randomization_ratio <0-100>]
         [--write-fadvise-dontneed]
         [--max-attr-len <bytes>]
@@ -161,6 +162,13 @@ Read Routing
     Set ``LIBRADOS_OPERATION_LOCALIZE_READS`` on read operations,
     preferring the closest replica.
 
+``--crush-location <loc>``
+    Set the client's CRUSH location for zone-aware reads (e.g.,
+    ``datacenter=zone-0``).  The location is applied via
+    ``crush_location`` in the librados configuration and influences
+    which OSD shard handles split reads on EC pools.  Has no effect
+    unless the cluster is configured for zone-aware placement.
+
 ``--offlen_randomization_ratio <n>``
     Percentage chance (0–100) that a read uses a randomized offset
     instead of reading from offset 0. Default: ``50``.

@@ -84,6 +84,7 @@ Mon
 - [x] common/options/global.yaml.in remove `mon_debug_no_require_X-2`
 - [x] mon/OSDMonitor.cc `create_initial`: adjust new `require_osd_release`, and add associated `mon_debug_no_require_X`
 - [x] mon/OSDMonitor.cc `preprocess_boot`: adjust "disallow boot of " condition to disallow X if `require_osd_release` < X-2.
+- [x] common/ceph_releases.cc `ceph_release_server_feature`: return `SERVER_X` for X, so that `preprocess_boot` refuses OSDs of earlier releases once `require_osd_release` is X (`unittest_features` fails until it does)
 - [x] mon/OSDMonitor.cc: adjust "osd require-osd-release" to (1) allow setting X, and (2) check that all mons *and* OSDs have X
 - [x] mon/MonCommands.h: adjust "osd require-osd-release" allows options to include X
 - [x] qa/workunits/cephtool/test.sh: adjust `require-osd-release` test

@@ -57,14 +57,22 @@ See :ref:`pg-autoscaler`.
 .. confval:: mon_delta_reset_interval
 .. confval:: osd_crush_chooseleaf_type
 .. confval:: osd_crush_initial_weight
+.. confval:: osd_pool_default_type
 .. confval:: osd_pool_default_crush_rule
 .. confval:: osd_pool_erasure_code_stripe_unit
 .. confval:: osd_pool_default_size
+.. confval:: osd_pool_default_num_zones
+.. confval:: osd_pool_default_zone_failure_domain
+.. confval:: osd_pool_default_osd_failure_domain
+.. confval:: osd_pool_default_root
+.. confval:: osd_pool_default_class
 .. confval:: osd_pool_default_min_size
 .. confval:: osd_pool_default_pg_num
 .. confval:: osd_pool_default_pgp_num
 .. confval:: osd_pool_default_pg_autoscale_mode
 .. confval:: osd_pool_default_flags
+.. confval:: osd_pool_default_flag_bulk
+.. confval:: osd_pool_default_crimson
 .. confval:: osd_max_pgls
 .. confval:: osd_min_pg_log_entries
 .. confval:: osd_max_pg_log_entries

@@ -741,6 +741,9 @@ The relevant erasure-code profile properties are as follows:
    a CRUSH MSR rule being created.
  * **crush-device-class**: the device class on which to place data [default:
    none, which means that all devices are used].
+ * **crush-zone-failure-domain**: the CRUSH bucket type of the zones of a pool
+   or rule over more than one zone [default:
+   :confval:`osd_pool_default_zone_failure_domain`].
  * **k** and **m** (and, for the ``lrc`` plugin, **l**): these determine the
    number of erasure-code shards, affecting the resulting CRUSH rule.
 
@@ -749,12 +752,47 @@ The relevant erasure-code profile properties are as follows:
 
 .. prompt:: bash $
 
-   ceph osd crush rule create-erasure {name} {profile-name}
+   ceph osd crush rule create-erasure {name} {profile-name} [{num_zones}]
+
+``num_zones`` (default ``1``) is the number of zones that the rule spans. A
+rule over more than one zone places the ``k+m`` shards in each of
+``num_zones`` buckets of the profile's ``crush-zone-failure-domain`` type, or
+of :confval:`osd_pool_default_zone_failure_domain` if the profile does not set
+one.
 
 .. note:: When creating a new pool, it is not necessary to create the rule
    explicitly. If only the erasure-code profile is specified and the rule
    argument is omitted, then Ceph will create the CRUSH rule automatically.
 
+Replicated CRUSH rules over zones
+---------------------------------
+
+Ceph generates a CRUSH rule for each replicated pool that is stretched over
+zones without being given one: by ``ceph osd pool create {pool-name}
+--num_zones 2``, by ``ceph osd pool set {pool-name} num_zones 2``, and by
+``ceph mon enable_stretch_mode``. The generated
+rule has this form, from the pool's ``replica`` and the
+:confval:`osd_pool_default_root`, :confval:`osd_pool_default_zone_failure_domain`,
+:confval:`osd_pool_default_osd_failure_domain` and
+:confval:`osd_pool_default_class` defaults, unless the command gives them::
+
+   rule {pool-name} {
+      id 2
+      type replicated
+      step take {root}
+      step choose firstn {num_zones} type {zone_failure_domain}
+      step chooseleaf firstn {replica} type {osd_failure_domain}
+      step emit
+   }
+
+If ``{osd_failure_domain}`` is ``osd``, the rule uses ``step choose firstn``
+instead of ``step chooseleaf firstn``.
+
+A rule shared by several pools, or a rule over three or more zones for an
+individual stretch pool (see :ref:`individual_stretch_pools`), is written by hand: get the
+CRUSH map with ``ceph osd getcrushmap``, decompile it with ``crushtool -d``, add
+a rule of the form above with an unused ``id``, compile it with ``crushtool
+-c`` and set it with ``ceph osd setcrushmap``.
 
 .. _rados-crush-msr-rules:
 
@@ -776,8 +814,9 @@ OSDs and clients are required to support the CRUSH_MSR feature bit
 Deleting rules
 --------------
 
-To delete rules that are not in use by pools, run a command of the following
-form:
+The Monitors delete a rule automatically when the last pool that uses it is
+deleted or given another rule, unless it is the default replicated rule. To
+delete a rule that no pool uses, run a command of the following form:
 
 .. prompt:: bash $
 

@@ -41,6 +41,62 @@ same time.
 	erasure-code-shec
 	erasure-code-clay
 
+.. _erasure-code-profile-lifecycle:
+
+Profile lifecycle
+=================
+
+An erasure code profile is deleted with the last pool that uses it.
+
+A profile is created in one of two ways:
+
+* explicitly, with ``ceph osd erasure-code-profile set`` (see below), before
+  creating the pool that uses it, or
+* automatically, when an erasure-coded pool is created without a profile name
+  (or with ``default``). The pool gets a profile of its own, named
+  ``{pool-name}-k{k}-m{m}``: a copy of the ``default`` profile with the
+  ``k``, ``m`` and CRUSH options given on the command line. For example, the
+  following command creates a profile named ``ecpool-k4-m2`` and a pool that
+  uses it:
+
+  .. prompt:: bash $
+
+     ceph osd pool create ecpool erasure --k 4 --m 2
+
+  If a profile of that name already exists with the same contents, it is used.
+  If it exists with other contents, the command fails: remove that profile, or
+  name it with ``--erasure_code_profile``.
+
+The ``default`` profile is a template that no pool uses. It holds what a pool
+gets when it names no profile, including its placement: ``crush-root``,
+``crush-zone-failure-domain``, ``crush-failure-domain`` and
+``crush-device-class`` (empty for no class). The profile always has these
+four: a key it lacks is filled in from :confval:`osd_pool_default_root`,
+:confval:`osd_pool_default_zone_failure_domain`,
+:confval:`osd_pool_default_osd_failure_domain` or
+:confval:`osd_pool_default_class` when the profile is created or set again.
+``ceph osd pool default set --k {k} --m {m}``
+changes its ``k`` and ``m`` and keeps its other keys. While the default pool
+type is ``erasure``, ``ceph osd pool default set`` writes ``--root``,
+``--zone_failure_domain``, ``--osd_failure_domain``, ``--class`` and
+``--no-class`` into it, instead of into the options. It refuses ``--k`` and
+``--m`` while the default pool type is ``replicated``, unless it is also given
+``--pool_type erasure``, which makes every later pool that names no type an
+erasure-coded pool.
+``ceph osd erasure-code-profile set default ...`` replaces the whole profile:
+with ``osd_pool_default_erasure_code_profile`` and the keys given, or with only
+the keys given if they name another plugin. Pools created before the change
+keep their own copies.
+
+When the last pool that uses a profile is deleted, the profile is deleted as
+well, however it was created. The ``default`` profile is never deleted in this
+way. Every erasure-coded pool gets a CRUSH rule of its own, named after the
+pool, unless it is given one; a CRUSH rule is likewise deleted when the last
+pool that uses it is deleted.
+
+To create another pool with a profile whose last pool has been deleted, set the
+profile again first.
+
 osd erasure-code-profile set
 ============================
 
@@ -51,7 +107,8 @@ To create a new erasure code profile::
              [{plugin=plugin}] \
              [{stripe_unit=stripe_unit}] \
              [{key=value} ...] \
-             [--force]
+             [--force] \
+             [--yes-i-really-mean-it]
 
 Where:
 
@@ -86,7 +143,10 @@ Where:
               created.  The stripe_width of a pool using this profile
               will be the number of data chunks multiplied by this
               stripe_unit. See :ref:`rados_ops_erasure_coding_optimizations`
-              for more information.
+              for more information. ``ceph osd pool create`` refuses a
+              profile whose stripe_unit is not a multiple of 4K unless
+              ``--yes-i-really-mean-it`` is given, and always for a pool
+              over zones.
 
 
 :Type: String
@@ -102,12 +162,22 @@ Where:
 
 ``--force``
 
-:Description: Override an existing profile by the same name, and allow
-              setting a non-4K-aligned stripe_unit. Overriding an existing
-              profile can be dangerous, and thus ``--yes-i-really-mean-it``
-              must be used as well.
+:Description: Allow setting a non-4K-aligned stripe_unit. With
+              ``--yes-i-really-mean-it``, also allow a ``blaum_roth``
+              profile whose ``w+1`` is not prime.
 
-:Type: String
+:Type: Boolean
+:Required: No.
+
+``--yes-i-really-mean-it``
+
+:Description: Override a profile that a pool uses. OSDs read a pool's
+              profile whenever they load its PGs, so changing the ``k``,
+              ``m`` or plugin of a profile that a pool uses corrupts the
+              pool's data. A profile that no pool uses, such as
+              ``default``, can be changed without it.
+
+:Type: Boolean
 :Required: No.
 
 osd erasure-code-profile rm
@@ -119,7 +189,12 @@ To remove an erasure code profile::
 
 If the profile is referenced by a pool, the deletion will fail.
 
-.. warning:: Removing an erasure code profile using ``osd erasure-code-profile rm`` does not automatically delete the associated CRUSH rule associated with the erasure code profile. It is recommended to manually remove the associated CRUSH rule using ``ceph osd crush rule remove {rule-name}`` to avoid unexpected behavior.
+It is rarely necessary to remove a profile, because a profile is deleted
+automatically with the last pool that uses it (see
+:ref:`erasure-code-profile-lifecycle`). Removing a profile that does not exist
+succeeds and reports that the profile does not exist.
+
+.. warning:: Removing an erasure code profile using ``osd erasure-code-profile rm`` does not delete a CRUSH rule created from it. A rule that a pool used is deleted with the last pool that uses it. To remove a rule that no pool has used, run ``ceph osd crush rule rm {rule-name}``.
 
 osd erasure-code-profile get
 ============================

@@ -62,12 +62,18 @@ profile can be displayed with this command:
 
 ::
 
+   crush-device-class=
+   crush-failure-domain=host
+   crush-root=default
+   crush-zone-failure-domain=datacenter
    k=2
    m=2
    plugin=isa
-   crush-failure-domain=host
    technique=reed_sol_van
 
+The ``crush-*`` keys are where erasure-coded pools that name no profile are
+placed (see :ref:`pool-creation-defaults`).
+
 .. note::
   The profile just displayed is for the *default* erasure-coded pool, not the
   *simplest* erasure-coded pool. These two pools are not the same:
@@ -252,6 +258,9 @@ unsupported combination of plugin and technique, such as CLAY, LRC, or
 SHEC, is blocked with an error message.
 
 The default stripe unit is 4K which works well for standard EC pools.
+A pool over two zones (``num_zones`` 2) is always optimized. Its stripe unit
+is 16K unless :confval:`osd_pool_erasure_code_stripe_unit` is set or its
+profile sets ``stripe_unit``.
 The stripe unit must be a multiple of 4 KiB for optimizations to be
 enabled; enabling the flag on a pool with any other stripe unit is
 rejected.