Using pg-upmap

In Luminous v12.2.z and later releases, there is a pg-upmap exception table in the OSDMap that allows the cluster to explicitly map specific PGs to specific OSDs. This allows the cluster to fine-tune the data distribution to, in most cases, uniformly distribute PGs across OSDs.

However, there is an important caveat when it comes to this new feature: it requires all clients to understand the new pg-upmap structure in the OSDMap.

Enabling

In order to use pg-upmap, the cluster cannot have any pre-Luminous clients. By default, new clusters enable the balancer module, which makes use of pg-upmap. If you want to use a different balancer or you want to make your own custom pg-upmap entries, you might want to turn off the balancer in order to avoid conflict:

ceph balancer off

To allow use of the new feature on an existing cluster, you must restrict the cluster to supporting only Luminous (and newer) clients. To do so, run the following command:

ceph osd set-require-min-compat-client luminous

This command will fail if any pre-Luminous clients or daemons are connected to the monitors. To see which client versions are in use, run the following command:

ceph features

Balancer module

The balancer module for ceph-mgr will automatically balance the number of PGs per OSD. See Balancer

Offline optimization

Upmap entries are updated with an offline optimizer that is built into osdmaptool.

  1. Grab the latest copy of your osdmap:

    ceph osd getmap -o om
    
  2. Run the optimizer:

    osdmaptool om --upmap out.txt [--upmap-pool <pool>] \
    [--upmap-max <max-optimizations>] \
    [--upmap-deviation <max-deviation>] \
    [--upmap-active]
    

    It is highly recommended that optimization be done for each pool individually, or for sets of similarly utilized pools. You can specify the --upmap-pool option multiple times. “Similarly utilized pools” means pools that are mapped to the same devices and that store the same kind of data (for example, RBD image pools are considered to be similarly utilized; an RGW index pool and an RGW data pool are not considered to be similarly utilized).

    The max-optimizations value determines the maximum number of upmap entries to identify. The default is 10 (as is the case with the ceph-mgr balancer module), but you should use a larger number if you are doing offline optimization. If it cannot find any additional changes to make (that is, if the pool distribution is perfect), it will stop early.

    The max-deviation value defaults to 5. If an OSD’s PG count varies from the computed target number by no more than this amount it will be considered perfect.

    The --upmap-active option simulates the behavior of the active balancer in upmap mode. It keeps cycling until the OSDs are balanced and reports how many rounds have occurred and how long each round takes. The elapsed time for rounds indicates the CPU load that ceph-mgr consumes when it computes the next optimization plan.

  3. Apply the changes:

    source out.txt
    

    In the above example, the proposed changes are written to the output file out.txt. The commands in this procedure are normal Ceph CLI commands that can be run in order to apply the changes to the cluster.

The above steps can be repeated as many times as necessary to achieve a perfect distribution of PGs for each set of pools.

To see some (gory) details about what the tool is doing, you can pass --debug-osd 10 to osdmaptool. To see even more details, pass --debug-crush 10 to osdmaptool.