PLEASE NOTE: This document applies to v1.6 version and not to the latest stable release v1.9
Ceph Shared Filesystem CRD
Rook allows creation and customization of shared filesystems through the custom resource definitions (CRDs). The following settings are available for Ceph filesystems.
NOTE: This sample requires at least 1 OSD per node, with each OSD located on 3 different nodes.
Each OSD must be located on a different node, because both of the defined pools set the
host and the
failureDomain can also be set to another location type (e.g.
rack), if it has been added as a
location in the Storage Selection Settings.
apiVersion: ceph.rook.io/v1 kind: CephFilesystem metadata: name: myfs namespace: rook-ceph spec: metadataPool: failureDomain: host replicated: size: 3 dataPools: - failureDomain: host replicated: size: 3 preserveFilesystemOnDelete: true metadataServer: activeCount: 1 activeStandby: true # A key/value list of annotations annotations: # key: value placement: # nodeAffinity: # requiredDuringSchedulingIgnoredDuringExecution: # nodeSelectorTerms: # - matchExpressions: # - key: role # operator: In # values: # - mds-node # tolerations: # - key: mds-node # operator: Exists # podAffinity: # podAntiAffinity: # topologySpreadConstraints: resources: # limits: # cpu: "500m" # memory: "1024Mi" # requests: # cpu: "500m" # memory: "1024Mi"
(These definitions can also be found in the
Erasure coded pools require the OSDs to use
bluestore for the configured
storeType. Additionally, erasure coded pools can only be used with
metadataPool must use a replicated pool.
NOTE: This sample requires at least 3 bluestore OSDs, with each OSD located on a different node.
The OSDs must be located on different nodes, because the
failureDomain will be set to
host by default, and the
erasureCoded chunk settings require at least 3 different OSDs (2
dataChunks + 1
apiVersion: ceph.rook.io/v1 kind: CephFilesystem metadata: name: myfs-ec namespace: rook-ceph spec: metadataPool: replicated: size: 3 dataPools: - erasureCoded: dataChunks: 2 codingChunks: 1 metadataServer: activeCount: 1 activeStandby: true
(These definitions can also be found in the
name: The name of the filesystem to create, which will be reflected in the pool and other resource names.
namespace: The namespace of the Rook cluster where the filesystem is created.
The pools allow all of the settings defined in the Pool CRD spec. For more details, see the Pool CRD settings. In the example above, there must be at least three hosts (size 3) and at least eight devices (6 data + 2 coding chunks) in the cluster.
metadataPool: The settings used to create the filesystem metadata pool. Must use replication.
dataPools: The settings to create the filesystem data pools. If multiple pools are specified, Rook will add the pools to the filesystem. Assigning users or files to a pool is left as an exercise for the reader with the CephFS documentation. The data pools can use replication or erasure coding. If erasure coding pools are specified, the cluster must be running with bluestore enabled on the OSDs.
preserveFilesystemOnDelete: If it is set to ‘true’ the filesystem will remain when the CephFilesystem resource is deleted. This is a security measure to avoid loss of data if the CephFilesystem resource is deleted accidentally. The default value is ‘false’. This option replaces
preservePoolsOnDeletewhich should no longer be set.
preservePoolsOnDelete: This option is replaced by the above
preserveFilesystemOnDelete. For backwards compatibility and upgradeability, if this is set to ‘true’, Rook will treat
preserveFilesystemOnDeleteas being set to ‘true’.
Metadata Server Settings
The metadata server settings correspond to the MDS daemon settings.
activeCount: The number of active MDS instances. As load increases, CephFS will automatically partition the filesystem across the MDS instances. Rook will create double the number of MDS instances as requested by the active count. The extra instances will be in standby mode for failover.
activeStandby: If true, the extra MDS instances will be in active standby mode and will keep a warm cache of the filesystem metadata for faster failover. The instances will be assigned by CephFS in failover pairs. If false, the extra MDS instances will all be on passive standby mode and will not maintain a warm cache of the metadata.
annotations: Key value pair list of annotations to add.
labels: Key value pair list of labels to add.
placement: The mds pods can be given standard Kubernetes placement restrictions with
podAntiAffinitysimilar to placement defined for daemons configured by the cluster CRD.
resources: Set resource requests/limits for the Filesystem MDS Pod(s), see MDS Resources Configuration Settings
priorityClassName: Set priority class name for the Filesystem MDS Pod(s)
MDS Resources Configuration Settings
The format of the resource requests/limits structure is the same as described in the Ceph Cluster CRD documentation.
If the memory resource limit is declared Rook will automatically set the MDS configuration
mds_cache_memory_limit. The configuration value is calculated with the aim that the actual MDS memory consumption remains consistent with the MDS pods’ resource declaration.
In order to provide the best possible experience running Ceph in containers, Rook internally recommends the memory for MDS daemons to be at least 4096MB. If a user configures a limit or request value that is too low, Rook will still run the pod(s) and print a warning to the operator log.