Skip to content
This repository has been archived by the owner on Oct 23, 2024. It is now read-only.

Latest commit

 

History

History
217 lines (174 loc) · 8.02 KB

resources.md

File metadata and controls

217 lines (174 loc) · 8.02 KB
Raw

Resources

Computable Resources

KUDO Cassandra by default requests 1.5 cpu and 4.5Gi of memory for each Cassandra node. The default limits are 2 cpu and 4.5Gi of memory.

Those requests and limits can be tuned using the following parameters:

  • NODE_CPU_MC
  • NODE_CPU_LIMIT_MC
  • NODE_MEM_MIB
  • NODE_MEM_LIMIT_MIB
  • PROMETHEUS_EXPORTER_CPU_MC
  • PROMETHEUS_EXPORTER_CPU_LIMIT_MC
  • PROMETHEUS_EXPORTER_MEM_MIB
  • PROMETHEUS_EXPORTER_MEM_LIMIT_MIB
  • RECOVERY_CONTROLLER_CPU_MC
  • RECOVERY_CONTROLLER_CPU_LIMIT_MC
  • RECOVERY_CONTROLLER_MEM_MIB
  • RECOVERY_CONTROLLER_MEM_LIMIT_MIB

Storage resources

By default, KUDO Cassandra uses 20GiB PV. This isn't recommended for production use. Please refer to production docs to see the storage and compute resources recommendations.

Resources per container

Cassandra container

resources:
  limits:
    cpu: 1
    memory: 4Gi
  requests:
    cpu: 1
    memory: 4Gi

Bootstrap init container

resources:
  limits:
    cpu: 200m
    memory: 256Mi
  requests:
    cpu: 100m
    memory: 128Mi

prometheus exporter sidecar

resources:
  limits:
    cpu: 1
    memory: 512Mi
  requests:
    cpu: 500m
    memory: 512Mi

cassandra-recovery controller pod

resources:
  limits:
    cpu: 200m
    memory: 256Mi
  requests:
    cpu: 50m
    memory: 50Mi

Kubernetes Objects

KUDO Cassandra is delivered with a specific set of features, which can be enabled if needed. enabling those features creates more objects in Kubernetes than default installation.

Let’s take a look at the resources created by default when doing a simple install:

$ kubectl kudo install cassandra
operator.kudo.dev/v1beta1/cassandra created
operatorversion.kudo.dev/v1beta1/cassandra-<version> created
instance.kudo.dev/v1beta1/cassandra-instance created
instance.kudo.dev/v1beta1/cassandra-instance ready

$ kubectl tree instance cassandra-instance
NAMESPACE  NAME                                                             READY  REASON  AGE
default    Instance/cassandra-instance                                      -              6m11s
default    ├─ConfigMap/cassandra-instance-cassandra-env-sh                  -              6m9s
default    ├─ConfigMap/cassandra-instance-cassandra-exporter-config-yml     -              6m9s
default    ├─ConfigMap/cassandra-instance-generate-cassandra-yaml           -              6m9s
default    ├─ConfigMap/cassandra-instance-generate-cqlshrc-sh               -              6m9s
default    ├─ConfigMap/cassandra-instance-generate-nodetool-ssl-properties  -              6m9s
default    ├─ConfigMap/cassandra-instance-generate-tls-artifacts-sh         -              6m9s
default    ├─ConfigMap/cassandra-instance-jvm-options                       -              6m9s
default    ├─ConfigMap/cassandra-instance-node-scripts                      -              6m9s
default    ├─ConfigMap/cassandra-instance-topology-lock                     -              6m9s
default    ├─PodDisruptionBudget/cassandra-instance-pdb                     -              6m9s
default    ├─Role/cassandra-instance-node-role                              -              6m10s
default    ├─Role/cassandra-instance-role                                   -              6m9s
default    ├─RoleBinding/cassandra-instance-binding                         -              6m9s
default    ├─RoleBinding/cassandra-instance-node-default-binding            -              6m10s
default    ├─Secret/cassandra-instance-tls-store-credentials                -              6m9s
default    ├─Service/cassandra-instance-svc                                 -              6m9s
default    ├─ServiceAccount/cassandra-instance-sa                           -              6m9s
default    ├─ServiceMonitor/cassandra-instance-monitor                      -              6m9s
default    └─StatefulSet/cassandra-instance-node                            -              6m9s
default      ├─ControllerRevision/cassandra-instance-node-659c89769d        -              2m24s
default      ├─Pod/cassandra-instance-node-0                                True           2m14s
default      ├─Pod/cassandra-instance-node-1                                True           94s
default      └─Pod/cassandra-instance-node-2                                True           53s

Statefulset

Statefulsets are designed to manage stateful workload in Kubernetes. KUDO Cassandra uses statefulsets. The operator by default uses OrderedReady pod management policy. This guarantees that pods are created sequentially, which means that when the Cassandra cluster is coming up, only one node starts at a time. Pod names are -node- starting from ordinal-index 0. For example a 3 node cluster created using KUDO Cassandra instance name cass-prod will have these pods:

$ kubectl get pods
NAME               READY   STATUS    RESTARTS   AGE
cass-prod-node-0   1/1     Running   0          101s
cass-prod-node-1   1/1     Running   0          49s
cass-prod-node-2   1/1     Running   0          8s

When a multi-datacenter configuration with NODE_TOPOLOGY is used, the pod names include the datacenter name as well. See multi-datacenter documentation.

Configmaps

KUDO Cassandra generates the configurable scripts and properties used in KUDO Cassandra operator as configmap objects.

PodDisruptionBudget

KUDO Cassandra limits the number of pods that may be down simultaneously. For Cassandra’s service to work without interruptions, especially when quorum-based applications are running on top of Cassandra, we need to guarantee that the number of replicas running is never brought below the number required for a quorum, even temporarily. Unlike a regular pod deletion, for the KUDO Cassandra pod eviction, the API server may reject the operation if the eviction would cause the disruption budget to be exceeded.

ServiceAccount / Role / RoleBinding

KUDO Cassandra creates one service account which is attached with two types of roles. The service account is -sa and then the Roles are:

<instance-name>-node-role
<instance-name>-role

The node role is used by KUDO Cassandra's recovery feature which uses the Kubernetes API to detect any deleted kubelets. It is also used to free up the KUDO Cassandra PVC Claim Ref if that is necessary. The generic role, <instance-name>-role is used by the Cassandra node bootstrap binary to update the topology-lock configmap so it has access to the configmaps

Secrets

KUDO Cassandra creates the TLS store credentials as a secret -tls-store-credentials. Those credentials are used as keystore/truststore credentials, when adding certificates to them. This is done when KUDO Cassandra's SSL feature is enabled.

Service

KUDO Cassandra creates a headless service -svc (ClusterIP type). This service is used for internal access from inside the Kubernetes cluster.

$ kubectl get svc
NAME                     TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)                               AGE
cassandra-instance-svc   ClusterIP   10.0.51.69   <none>        7000/TCP,7001/TCP,9042/TCP,7200/TCP   26m

The service exposes the storage port on 7000 by default and on 7001 if SSL is enabled. The native transport defaults to port 9042, and metrics are exposed via port 7200 by default.

The RPC port is disabled by default and can be enabled using the parameter START_RPC=true, which will expose the RPC port on 9160 by default. All above information can be exposed via custom ports using the parameters:

STORAGE_PORT
SSL_STORAGE_PORT
NATIVE_TRANSPORT_PORT
RPC_PORT

ServiceMonitor

KUDO Cassandra integrates with prometheus-operator by default. The ServiceMonitor uses the labels kudo.dev/servicemonitor and kudo.dev/instance to discover the Cassandra pods.

To disable monitoring, users can start KUDO Cassandra with the parameter PROMETHEUS_EXPORTER_ENABLED=false. Read more information about monitoring in the Monitoring section