]> git-server-git.apps.pok.os.sepia.ceph.com Git - ceph.git/commitdiff
doc/rados/operations: adjust docs a bit 33092/head
authorSage Weil <sage@redhat.com>
Thu, 6 Feb 2020 13:09:29 +0000 (07:09 -0600)
committerSage Weil <sage@redhat.com>
Thu, 6 Feb 2020 13:09:29 +0000 (07:09 -0600)
Signed-off-by: Sage Weil <sage@redhat.com>
doc/man/8/ceph.rst
doc/rados/operations/pools.rst

index 7e125b96b91f24fe4c458baf1e54b586fd5e8bc5..7c98629637f01506ba778910d9e2fd1bc8c4e8f5 100644 (file)
@@ -1036,7 +1036,7 @@ Subcommand ``create`` creates pool.
 Usage::
 
        ceph osd pool create <poolname> {<int[0-]>} {<int[0-]>} {replicated|erasure}
-       {<erasure_code_profile>} {<rule>} {<int>}
+       {<erasure_code_profile>} {<rule>} {<int>} {--autoscale-mode=<on,off,warn>}
 
 Subcommand ``delete`` deletes pool.
 
index 8aa6f378b69c351219830c3ef1e26ab4446dc0f7..70dca8c75dc619bfc0d813531cd3f47b1281f140 100644 (file)
@@ -61,7 +61,7 @@ To create a pool, execute::
        ceph osd pool create {pool-name} [{pg-num} [{pgp-num}]] [replicated] \
              [crush-rule-name] [expected-num-objects]
        ceph osd pool create {pool-name} [{pg-num} [{pgp-num}]]   erasure \
-             [erasure-code-profile] [crush-rule-name] [expected_num_objects]
+             [erasure-code-profile] [crush-rule-name] [expected_num_objects] [--autoscale-mode=<on,off,warn>]
 
 Where:
 
@@ -131,15 +131,15 @@ Where:
 :Type: String
 :Required: No.
 
-When you create a pool, set the number of placement groups to a reasonable value
-(e.g., ``100``). Consider the total number of placement groups per OSD too.
-Placement groups are computationally expensive, so performance will degrade when
-you have many pools with many placement groups (e.g., 50 pools with 100
-placement groups each). The point of diminishing returns depends upon the power
-of the OSD host.
+``--autoscale-mode=<on,off,warn>``
 
-See `Placement Groups`_ for details on calculating an appropriate number of
-placement groups for your pool.
+:Description: Autoscale mode
+
+:Type: String
+:Required: No.
+:Default:  The default behavior is controlled by the ``osd pool default pg autoscale mode`` option.
+
+If you set the autoscale mode to ``on`` or ``warn``, you can let the system autotune or recommend changes to the number of placement groups in your pool based on actual usage.  If you leave it off, then you should refer to `Placement Groups`_ for more information.
 
 .. _Placement Groups: ../placement-groups