ceph/doc/rados/operations/upmap.rst

   1 .. _upmap:
   2
   3 Using pg-upmap
   4 ==============
   5
   6 In Luminous v12.2.z and later releases, there is a *pg-upmap* exception table
   7 in the OSDMap that allows the cluster to explicitly map specific PGs to
   8 specific OSDs. This allows the cluster to fine-tune the data distribution to,
   9 in most cases, uniformly distribute PGs across OSDs.
  10
  11 However, there is an important caveat when it comes to this new feature: it
  12 requires all clients to understand the new *pg-upmap* structure in the OSDMap.
  13
  14 Enabling
  15 --------
  16
  17 In order to use ``pg-upmap``, the cluster cannot have any pre-Luminous clients.
  18 By default, new clusters enable the *balancer module*, which makes use of
  19 ``pg-upmap``. If you want to use a different balancer or you want to make your
  20 own custom ``pg-upmap`` entries, you might want to turn off the balancer in
  21 order to avoid conflict:
  22
  23 .. prompt:: bash $
  24
  25    ceph balancer off
  26
  27 To allow use of the new feature on an existing cluster, you must restrict the
  28 cluster to supporting only Luminous (and newer) clients.  To do so, run the
  29 following command:
  30
  31 .. prompt:: bash $
  32
  33    ceph osd set-require-min-compat-client luminous
  34
  35 This command will fail if any pre-Luminous clients or daemons are connected to
  36 the monitors. To see which client versions are in use, run the following
  37 command:
  38
  39 .. prompt:: bash $
  40
  41    ceph features
  42
  43 Balancer module
  44 ---------------
  45
  46 The `balancer` module for ``ceph-mgr`` will automatically balance the number of
  47 PGs per OSD. See :ref:`balancer`
  48
  49 Offline optimization
  50 --------------------
  51
  52 Upmap entries are updated with an offline optimizer that is built into
  53 ``osdmaptool``.
  54
  55 #. Grab the latest copy of your osdmap:
  56
  57    .. prompt:: bash $
  58
  59       ceph osd getmap -o om
  60
  61 #. Run the optimizer:
  62
  63    .. prompt:: bash $
  64
  65       osdmaptool om --upmap out.txt [--upmap-pool <pool>] \
  66       [--upmap-max <max-optimizations>] \
  67       [--upmap-deviation <max-deviation>] \
  68       [--upmap-active]
  69
  70    It is highly recommended that optimization be done for each pool
  71    individually, or for sets of similarly utilized pools. You can specify the
  72    ``--upmap-pool`` option multiple times. "Similarly utilized pools" means
  73    pools that are mapped to the same devices and that store the same kind of
  74    data (for example, RBD image pools are considered to be similarly utilized;
  75    an RGW index pool and an RGW data pool are not considered to be similarly
  76    utilized).
  77
  78    The ``max-optimizations`` value determines the maximum number of upmap
  79    entries to identify. The default is `10` (as is the case with the
  80    ``ceph-mgr`` balancer module), but you should use a larger number if you are
  81    doing offline optimization.  If it cannot find any additional changes to
  82    make (that is, if the pool distribution is perfect), it will stop early.
  83
  84    The ``max-deviation`` value defaults to `5`. If an OSD's PG count varies
  85    from the computed target number by no more than this amount it will be
  86    considered perfect.
  87
  88    The ``--upmap-active`` option simulates the behavior of the active balancer
  89    in upmap mode. It keeps cycling until the OSDs are balanced and reports how
  90    many rounds have occurred and how long each round takes. The elapsed time
  91    for rounds indicates the CPU load that ``ceph-mgr`` consumes when it computes
  92    the next optimization plan.
  93
  94 #. Apply the changes:
  95
  96    .. prompt:: bash $
  97
  98       source out.txt
  99
 100    In the above example, the proposed changes are written to the output file
 101    ``out.txt``. The commands in this procedure are normal Ceph CLI commands
 102    that can be run in order to apply the changes to the cluster.
 103
 104 The above steps can be repeated as many times as necessary to achieve a perfect
 105 distribution of PGs for each set of pools.
 106
 107 To see some (gory) details about what the tool is doing, you can pass
 108 ``--debug-osd 10`` to ``osdmaptool``. To see even more details, pass
 109 ``--debug-crush 10`` to ``osdmaptool``.