Skip to content

Latest commit

 

History

History
388 lines (358 loc) · 22.4 KB

edgefs-cluster-crd.md

File metadata and controls

388 lines (358 loc) · 22.4 KB
Raw
title weight indent
EdgeFS Cluster CRD
4100
true

EdgeFS Cluster CRD

Rook allows creation and customization of storage clusters through the custom resource definitions (CRDs).

Sample

To get you started, here is a simple example of a CRD to configure a EdgeFS cluster with just one local per-host directory /data:

apiVersion: edgefs.rook.io/v1alpha1
kind: Cluster
metadata:
  name: rook-edgefs
  namespace: rook-edgefs
spec:
  edgefsImageName: edgefs/edgefs:latest
  serviceAccount: rook-edgefs-cluster
  dataDirHostPath: /data
  storage:
    useAllNodes: true

or if you have raw block devices provisioned, it can dynamically detect, format and utilize all raw devices on all nodes with simple CRD as below:

apiVersion: edgefs.rook.io/v1alpha1
kind: Cluster
metadata:
  name: rook-edgefs
  namespace: rook-edgefs
spec:
  edgefsImageName: edgefs/edgefs:latest
  serviceAccount: rook-edgefs-cluster
  dataDirHostPath: /data
  storage:
    useAllNodes: true
    useAllDevices: true

In addition to the CRD, you will also need to create a namespace, role, and role binding as seen in the common cluster resources below.

Settings

Settings can be specified at the global level to apply to the cluster as a whole, while other settings can be specified at more fine-grained levels, e.g. individual nodes. If any setting is unspecified, a suitable default will be used automatically.

Cluster metadata

  • name: The name that will be used internally for the EdgeFS cluster. Most commonly the name is the same as the namespace since multiple clusters are not supported in the same namespace.
  • namespace: The Kubernetes namespace that will be created for the Rook cluster. The services, pods, and other resources created by the operator will be added to this namespace. The common scenario is to create a single Rook cluster. If multiple clusters are created, they must not have conflicting devices or host paths.
  • edgefsImageName: EdgeFS image to use. If not specified then edgefs/edgefs:latest is used. We recommend to specify particular image version for production use, for example edgefs/edgefs:1.0.0.

Cluster Settings

  • dataDirHostPath: The path on the host (hostPath) where config and data should be stored for each of the services. If the directory does not exist, it will be created. Because this directory persists on the host, it will remain after pods are deleted. If storage settings not provided then provisioned hostPath will also be used as a storage device for Target pods (automatic provisioning via rtlfs).
    • On Minikube environments, use /data/rook. Minikube boots into a tmpfs but it provides some directories where files can be persisted across reboots. Using one of these directories will ensure that Rook's data and configuration files are persisted and that enough storage space is available.
    • WARNING: For test scenarios, if you delete a cluster and start a new cluster on the same hosts, the path used by dataDirHostPath must be deleted. Otherwise, stale information and other config will remain from the previous cluster and the new target will fail to start. If this value is empty, each pod will get an ephemeral directory to store their config files that is tied to the lifetime of the pod running on that node. More details can be found in the Kubernetes empty dir docs.
  • dataVolumeSize: Alternative to dataDirHostPath. If defined then Cluster CRD operator will disregard dataDirHostPath setting and instead will automatically claim persistent volume. If storage settings not provided then provisioned volume will also be used as a storage device for Target pods (automatic provisioning via rtlfs).
  • network: If defined then host network will be enabled for the cluster and services. This is optional and if not defined then eth0 will be used to construct cluster bucket network.
    • serverIfName: Specifies data daemon networking interface name. If not defined then eth0 is assumed.
    • brokerIfName: Specifies broker daemon networking interface name. If not defined then eth0 is assumed.
  • devicesResurrectMode: When enabled, this mode attempts to recreate cluster based on previous CRD definition. If this flag set to one of the parameters, then operator will only adjust networking. Often used when clean up of old devices is needed. Only applicable when used with dataDirHostPath.
    • restore: Attempt to restart and restore previously enabled cluster CRD.
    • restoreZap: Attempt to re-initialize previously selected devices prior to restore. By default cluster assumes that selected devices have no logical partitions and considered empty.
    • restoreZapWait: Attempt to cleanup previously selected devices and wait for cluster delete. This is useful when clean up of old devices is needed.
  • serviceAccount: The service account under which the EdgeFS pods will run that will give access to ConfigMaps in the cluster's namespace. If not set, the default of rook-edgefs-cluster will be used.
  • placement: placement configuration settings
  • resources: resources configuration settings
  • storage: Storage selection and configuration that will be used across the cluster. Note that these settings can be overridden for specific nodes.
    • useAllNodes: true or false, indicating if all nodes in the cluster should be used for storage according to the cluster level storage selection and configuration values. If individual nodes are specified under the nodes field below, then useAllNodes must be set to false.
    • nodes: Names of individual nodes in the cluster that should have their storage included in accordance with either the cluster level configuration specified above or any node specific overrides described in the next section below. useAllNodes must be set to false to use specific nodes and their config.
    • storage selection settings
    • storage configuration settings
  • skipHostPrepare: By default all nodes selected for EdgeFS deployment will be automatically configured via preparation jobs. If this option set to true node configuration will be skipped.

Node Updates

Nodes can be added and removed over time by updating the Cluster CRD, for example with kubectl -n rook-edgefs edit cluster.edgefs.rook.io rook-edgefs. This will bring up your default text editor and allow you to add and remove storage nodes from the cluster. This feature is only available when useAllNodes has been set to false.

Node Settings

In addition to the cluster level settings specified above, each individual node can also specify configuration to override the cluster level settings and defaults. If a node does not specify any configuration then it will inherit the cluster level settings.

Storage Selection Settings

Below are the settings available, both at the cluster and individual node level, for selecting which storage resources will be included in the cluster.

  • useAllDevices: true or false, indicating whether all devices found on nodes in the cluster should be automatically consumed by Targets. This is recommended for controlled environments where you will not risk formatting of devices with existing data. When true, all devices will be used except those with partitions created or a local filesystem. This can be overridden by deviceFilter.
  • deviceFilter: A regular expression that allows selection of devices to be consumed by target. If individual devices have been specified for a node then this filter will be ignored. This field uses golang regular expression syntax. For example:
    • sdb: Only selects the sdb device if found
    • ^sd.: Selects all devices starting with sd
    • ^sd[a-d]: Selects devices starting with sda, sdb, sdc, and sdd if found
    • ^s: Selects all devices that start with s
    • ^[^r]: Selects all devices that do not start with r
  • devices: A list of individual device names belonging to this node to include in the storage cluster. Mixing of devices and directories on the same node isn't supported.
    • name: The name of the device (e.g., sda).
    • config: Device-specific config settings. See the config settings below.
  • directories: A list of directory paths on the nodes that will be included in the storage cluster. Note that using two directories on the same physical device can cause a negative performance impact. Mixing of devices and directories on the same node isn't supported. Since EdgeFS is leveraging StatefulSet, directories can only be defined at cluster level.
    • path: The path on disk of the directory (e.g., /rook/storage-dir).
    • config: Directory-specific config settings. See the config settings below.

Storage Configuration Settings

The following storage selection settings are specific to EdgeFS and do not apply to other backends. All variables are key-value pairs represented as strings. While EdgeFS supports multiple backends, it is not recommended to mix them within same cluster. In case of devices (physical or emulated raw disks), EdgeFS will automatically use rtrd backend. In all other cases rtlfs (local file system) will be used. IMPORTANT Keys needs to be case-sensitive and values has to be provided as strings.

  • useMetadataOffload: Dynamically detect appropriate SSD/NVMe device to use for the metadata on each node. Performance can be improved by using a low latency device as the metadata device, while other spinning platter (HDD) devices on a node are used to store data. Typical and recommended proportion is in range of 1:1 - 1:6. Default is false. Applicable only to rtrd.
  • useMetadataMask: Defines what parts of metadata needs to be stored on offloaded devices. Default is 0x7d, offload all except second level manifests. For maximum performance, when you have enough SSD/NVMe capacity provisioned, set it to 0xff, i.e. all metadata. Applicable only to rtrd.
  • useBCache: When useMetadataOffload is true, enable use of BCache. Default is false. Applicable only to rtrd and when host has "bcache" kernel module preloaded.
  • useBCacheWB: When useMetadataOffload and useBCache is true, this option can enable use of BCache write-back cache. By default BCache only used as read cache in front of HDD. Applicable only to rtrd.
  • useAllSSD: When set to true, only SSD/NVMe non rotational devices will be used. Default is false and if useMetadataOffload not defined then only rotational devices (HDDs) will be picked up during node provisioning phase.
  • rtrdPLevelOverride: In case of large devices, it will be automatically partitioned into smaller parts around 500GB each. This option allows partitioning number override. Default is automatic. Typical and recommended range is 1 - 32. Applicable only to rtrd.
  • rtVerifyChid: Verify transfered or read payload. Payload can be data or metadata chunk of flexible size between 4K and 8MB. EdgeFS uses SHA-3 variant to crypthografically sign each chunk and uses it for self validation, self healing and FlexHash addressing. In case of low CPU systems verification after networking transfer prior to write can be disabled by setting this parameter to 0. In case of high CPU systems, verficication after read but before networking transfer can be enabled by setting this parameter to 2. Default is 1, i.e. verify after networking transfer only. Setting it to 0 may improve CPU utilization at the cost of reduced availability. However, for objects with 3 or more replicas, availability isn't going to be visibly affected.
  • lmdbPageSize: Defines default LMDB page size in bytes. Default is 16384. For capacity (all HDD) or hybrid (HDD/SSD) systems consider to increase this value to 32768 to achive higher throughput performance. For all SSD and small database workloads, consider to decrease this to 8192 to achive lower latency and higher IOPS. Please be advised that smaller values MAY cause fragmentation. Acceptable values are 4096, 8192, 16384 and 32768.
  • sync: Defines default behavior of write operations at device or directory level. Acceptible values are 0, 1 (default), 2, 3.
    • 0: No syncing will happen. Highest performance possible and good for HPC scratch types of deployments. This option will still sustain crash of pods or software bugs. It will not sustain server power loss an may cause node / device level inconsistency.
    • 1: Default method. Will guarantee node / device consistency in case of power loss with reduced durability.
    • 2: Provides better durability in case of power loss at the cost of extra metadata syncing.
    • 3: Most durable and reliable option at the cost of significant performance impact.

Placement Configuration Settings

Placement configuration for the cluster services. It includes the following keys: mgr, target and all. Each service will have its placement configuration generated by merging the generic configuration under all with the most specific one (which will override any attributes).

A Placement configuration is specified (according to the Kubernetes PodSpec) as:

The mgr pod does not allow Pod affinity or anti-affinity. This is because of the mgrs having built-in anti-affinity with each other through the operator. The operator chooses which nodes are to run a mgr on. Each mgr is then tied to a node with a node selector using a hostname.

Cluster-wide Resources Configuration Settings

Resources should be specified so that the rook components are handled after Kubernetes Pod Quality of Service classes. This allows to keep rook components running when for example a node runs out of memory and the rook components are not killed depending on their Quality of Service class.

You can set resource requests/limits for rook components through the Resource Requirements/Limits structure in the following keys:

  • mgr: Set resource requests/limits for Mgrs.
  • target: Set resource requests/limits for Targets.

Resource Requirements/Limits

For more information on resource requests/limits see the official Kubernetes documentation: Kubernetes - Managing Compute Resources for Containers

  • requests: Requests for cpu or memory.
    • cpu: Request for CPU (example: one CPU core 1, 50% of one CPU core 500m).
    • memory: Limit for Memory (example: one gigabyte of memory 1Gi, half a gigabyte of memory 512Mi).
  • limits: Limits for cpu or memory.
    • cpu: Limit for CPU (example: one CPU core 1, 50% of one CPU core 500m).
    • memory: Limit for Memory (example: one gigabyte of memory 1Gi, half a gigabyte of memory 512Mi).

Kubernetes node labeling and selection

By default each Kubernetes node, available to deploy EdgeFS over it, will be treated as "target" EdgeFS instance. But cluster administrator able to label node as "gateway" node, such node will have no devices prepared for EdgeFS and will be used as EdgeFS dedicated service node. To mark a node as a "gateway", the administrator can add a specific label to a node. Label format: <edgefs-namespace>-nodetype=gateway, where edgefs-namespace is current namespace for EdgeFS cluster deployment, by default is rook-edgefs Example: kubectl label node "k8s node name" rook-edgefs-nodetype=gateway

Samples

Here are several samples for configuring EdgeFS clusters. Each of the samples must also include the namespace and corresponding access granted for management by the EdgeFS operator. See the common cluster resources below.

Storage configuration: All devices, All SSD/NVMes.

apiVersion: edgefs.rook.io/v1alpha1
kind: Cluster
metadata:
  name: rook-edgefs
  namespace: rook-edgefs
spec:
  edgefsImageName: edgefs/edgefs:latest
  dataDirHostPath: /var/lib/rook
  serviceAccount: rook-edgefs-cluster
  # cluster level storage configuration and selection
  storage:
    useAllNodes: true
    useAllDevices: true
    deviceFilter:
    location:
    config:
      useAllSSD: true

Storage Configuration: Specific devices

Individual nodes and their config can be specified so that only the named nodes below will be used as storage resources. Each node's 'name' field should match their 'kubernetes.io/hostname' label.

apiVersion: edgefs.rook.io/v1alpha1
kind: Cluster
metadata:
  name: rook-edgefs
  namespace: rook-edgefs
spec:
  edgefsImageName: edgefs/edgefs:latest
  dataDirHostPath: /var/lib/rook
  serviceAccount: rook-edgefs-cluster
  # cluster level storage configuration and selection
  storage:
    useAllNodes: false
    useAllDevices: false
    deviceFilter:
    location:
    config:
      rtVerifyChid: "0"
    nodes:
    - name: "172.17.4.201"
      devices:             # specific devices to use for storage can be specified for each node
      - name: "sdb"
      - name: "sdc"
      config:         # configuration can be specified at the node level which overrides the cluster level config
        rtrdPLevelOverride: 8
    - name: "172.17.4.301"
      deviceFilter: "^sd."

Node Affinity

To control where various services will be scheduled by Kubernetes, use the placement configuration sections below. The example under 'all' would have all services scheduled on Kubernetes nodes labeled with 'role=storage' and tolerate taints with a key of 'storage-node'.

apiVersion: edgefs.rook.io/v1alpha1
kind: Cluster
metadata:
  name: rook-edgefs
  namespace: rook-edgefs
spec:
  edgefsImageName: edgefs/edgefs:latest
  dataDirHostPath: /var/lib/rook
  serviceAccount: rook-edgefs-cluster
  placement:
    all:
      nodeAffinity:
        requiredDuringSchedulingIgnoredDuringExecution:
          nodeSelectorTerms:
          - matchExpressions:
            - key: role
              operator: In
              values:
              - storage-node
      tolerations:
      - key: storage-node
        operator: Exists
    mgr:
      nodeAffinity:
      tolerations:
    target:
      nodeAffinity:
      tolerations:

Resource requests/Limits

To control how many resources the rook components can request/use, you can set requests and limits in Kubernetes for them. You can override these requests/limits for Targts per node when using useAllNodes: false in the node item in the nodes list.

apiVersion: edgefs.rook.io/v1alpha1
kind: Cluster
metadata:
  name: rook-edgefs
  namespace: rook-edgefs
spec:
  edgefsImageName: edgefs/edgefs:latest
  dataDirHostPath: /var/lib/rook
  serviceAccount: rook-edgefs-cluster
  # cluster level resource requests/limits configuration
  resources:
  storage:
    useAllNodes: false
    nodes:
    - name: "172.17.4.201"
      resources:
        limits:
          cpu: "2"
          memory: "4096Mi"
        requests:
          cpu: "2"
          memory: "4096Mi"

Common Cluster Resources

Each EdgeFS cluster must be created in a namespace and also give access to the Rook operator to manage the cluster in the namespace. Creating the namespace and these controls must be added to each of the examples previously shown.

apiVersion: v1
kind: Namespace
metadata:
  name: rook-edgefs
---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: rook-edgefs-cluster
  namespace: rook-edgefs
---
kind: Role
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: rook-edgefs-cluster
  namespace: rook-edgefs
rules:
- apiGroups: [""]
  resources: ["configmaps"]
  verbs: [ "get", "list", "watch", "create", "update", "delete" ]
- apiGroups: [""]
  resources: ["pods"]
  verbs: [ "get", "list" ]
---
# Allow the operator to create resources in this cluster's namespace
kind: RoleBinding
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: rook-edgefs-cluster-mgmt
  namespace: rook-edgefs
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: rook-edgefs-cluster-mgmt
subjects:
- kind: ServiceAccount
  name: rook-edgefs-system
  namespace: rook-edgefs-system
---
# Allow the pods in this namespace to work with configmaps
kind: RoleBinding
apiVersion: rbac.authorization.k8s.io/v1beta1
metadata:
  name: rook-edgefs-cluster
  namespace: rook-edgefs
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: rook-edgefs-cluster
subjects:
- kind: ServiceAccount
  name: rook-edgefs-cluster
  namespace: rook-edgefs
---
apiVersion: extensions/v1beta1
kind: PodSecurityPolicy
metadata:
  name: privileged
spec:
  fsGroup:
    rule: RunAsAny
  privileged: true
  runAsUser:
    rule: RunAsAny
  seLinux:
    rule: RunAsAny
  supplementalGroups:
    rule: RunAsAny
  volumes:
  - '*'
  allowedCapabilities:
  - '*'
  hostPID: true
  hostIPC: true
  hostNetwork: false
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: privileged-psp-user
rules:
- apiGroups:
  - apps
  resources:
  - podsecuritypolicies
  resourceNames:
  - privileged
  verbs:
  - use
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: rook-edgefs-system-psp
  namespace: rook-edgefs
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: privileged-psp-user
subjects:
- kind: ServiceAccount
  name: rook-edgefs-system
  namespace: rook-edgefs-system