Uploaded image for project: 'OpenStack Strategy'
  1. OpenStack Strategy
  2. RHOSSTRAT-1184

Ceph Erasure Coded Pools support for Storage

XMLWordPrintable

    • Icon: Feature Feature
    • Resolution: Unresolved
    • Icon: Major Major
    • None
    • None
    • Cinder, Glance, Manila
    • None
    • Not Selected
    • False
    • False
    • Hide

      None

      Show
      None
    • 0
    • 0
    • 100% To Do, 0% In Progress, 0% Done
    • rhos-storage

      Feature Overview 

      Add RBD EC pool support for Cinder, Glance and Manila

      Goals 

      • RHCSv9 greatly improves EC support coverage and performance. Given the potential cost reduction it can provide we should add support for Manila i.e CephFS native & CephFS-NFS and RBD for Cinder & Glance
      • Today customers are limited to replicated pools which has a bad impact in terms of cost since you need 2 or 3 time the usable capacity.

      Replicated pools should remain the default recommendation.

      Requirements 

      Requirement Notes isMVP?
      How to configure   yes
      Limitations documented   yes
      N+M recommendations   yes
           

       

      Support for the following services:

      • Cinder volume
      • Cinder backup
      • Glance
      • Manila CephFS

       

      Done - Acceptance Criteria

      • EC is configurable and passes the same tests as replicated pools
      • Stress test?
      • Perf/Scale test?

      Use Cases:
      As an administrator i would like to benefit from recent RHCS improvement on EC in order to reduce the cost of purchase and operation of my cluster.

      Out of Scope:
      Limited to openstack drivers, EC itself is covered by the ceph team.

       

      Documentation Considerations 

      Customer facing documentation for RBD EC pool use in Cinder, Glance and Manila

       

      Questions to Answer

      1. Does this feature track the Nova use case:  No, there will be a separate feature
      2. For manila it should matter if using native CephFS or CephFS-NFS
      3. Can we have heterogeneous deployments e.g
        1. Glance EC & Cinder replicated.
        2. Glance EC, Nova replicated.
        3. One cinder backend replicated, one EC
        4. Etc...
      4. Get performance data from the ceph team (compare vs replicated on same I/O pattern)
      5. Do we need to ask the P&S team to run a stress test?
      6. Can you transition from replicated to EC or vise versa, likely not so document it.

      Background and Strategic Fit (Initial completion while in Refinement status):
      Provide any additional context is needed to frame the feature.
      <your text here>

      Customer Considerations {}{}(Initial completion while in Refinement status):
      Provide any additional customer-specific considerations that must be made when designing and delivering the Feature.
      <your text here>

       

      EPIC Refinement Considerations:{}

      • Consider adding a Cinder Backup Epic separately

              rh-gs-gcharot Gregory Charot
              pgrist@redhat.com Paul Grist
              Gregory Charot Gregory Charot
              Edu Alcaniz Edu Alcaniz
              rhos-storage
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

                Created:
                Updated: