Enable specifying target size ratio for block, file and object pools #2979
Conversation
malayparida2000
force-pushed
the
target_size_ratio
branch
2 times, most recently
from
7b1b066 to
d1fe711
Compare
Although currently we support specifying only the target size ratio, this field can be used to support more fields in the future. Signed-off-by: Malay Kumar Parida <[email protected]>
malayparida2000
force-pushed
the
target_size_ratio
branch
from
d1fe711 to
c4bac83
Compare
|
/cc @iamniting @travisn |
| //lint:ignore ST1017 required to compare it directly | ||
| if "data" == poolType { | ||
| crs.TargetSizeRatio = .49 | ||
| if poolType == "data" { |
There was a problem hiding this comment.
Can you pls create const variables for block, file etc and use them at all places?
| var definedSize float64 | ||
|
|
||
| switch storageType { | ||
| case "block": |
There was a problem hiding this comment.
dont we need a case for the nfs and the API changes as well?
There was a problem hiding this comment.
Acc to the design doc discussion we don't need to expose for nfs block pool or replica-1 block pool.
There was a problem hiding this comment.
https://docs.google.com/document/d/1BOag-Hm9vpmYnpW76MstJXBdDCyAlYQf0rexpdKZnCA/edit?disco=AAABUdNdABs
But just to confirm again, @travisn do we need to
- expose a field for the nfs block pool?
- use the same value in the nfs blockpool as the default block pool?
- nfs block pool should stay at the default 0.49(current behaviour)
There was a problem hiding this comment.
For NFS, are you referring to the .nfs CephBlockPool? That is just a metadata pool, so it doesn't need the targetSizeRatio to be set. NFS uses a data pool from CephFS.
|
/test ocs-operator-bundle-e2e-aws |
Configuring the target size ratio enables Ceph to adjust PGs based on the anticipated usage of the pools. Currently all the dataPools (RBD/ CephFS/object) have a target_size_ratio of 0.49. Having same ratios for all data Pools causes under allocation of PGs for some pools & over allocation for others. According to the expected usage of the pools, the target size ratio can be set per pool. Signed-off-by: Malay Kumar Parida <[email protected]>
malayparida2000
force-pushed
the
target_size_ratio
branch
from
c4bac83 to
509637c
Compare
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: malayparida2000 The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
| @@ -199,7 +199,7 @@ func (o *ocsCephBlockPools) reconcileNFSCephBlockPool(r *StorageClusterReconcile | |||
| cephBlockPool.Spec.PoolSpec.DeviceClass = storageCluster.Status.DefaultCephDeviceClass | |||
| cephBlockPool.Spec.EnableCrushUpdates = true | |||
| cephBlockPool.Spec.PoolSpec.FailureDomain = getFailureDomain(storageCluster) | |||
| cephBlockPool.Spec.PoolSpec.Replicated = generateCephReplicatedSpec(storageCluster, "data") | |||
| cephBlockPool.Spec.PoolSpec.Replicated = generateCephReplicatedSpec(storageCluster, poolTypeData, storageTypeBlock) | |||
There was a problem hiding this comment.
dont we need a separate one here for nfs as it was earlier?
There was a problem hiding this comment.
Have asked Travis on what we need to do for nfs here
#2979 (comment)
|
|
||
| switch storageType { | ||
| case storageTypeBlock: | ||
| definedRatio = initData.Spec.ManagedResources.CephBlockPools.PoolSpec.Replicated.TargetSizeRatio |
There was a problem hiding this comment.
Actually, the PoolSpec.TargetSizeRatio was deprecated in rook a long time ago, but we have kept it for backward compatibility. Instead, let's use the parameters:
| definedRatio = initData.Spec.ManagedResources.CephBlockPools.PoolSpec.Replicated.TargetSizeRatio | |
| definedRatio = initData.Spec.ManagedResources.CephBlockPools.PoolSpec.Parameters["target_size_ratio"] |
Ref-https://issues.redhat.com/browse/RHSTOR-5690
Configuring the target size ratio enables Ceph to adjust PGs based on the anticipated usage of the pools. Currently all the dataPools (RBD/CephFS/object) have a target_size_ratio of 0.49. Having same ratios for all data Pools causes under allocation of PGs for some pools & over allocation for others. According to the expected usage of the pools, the target size ratio can be set per pool.