Troubleshoot K8ssandra Operator

Troubleshooting tips for K8ssandra Operator users.

The best place to start when troubleshooting a K8ssandra cluster deployment is its status. The status of a K8ssandra cluster reports useful information about each of its components (CassandraDatacenter, Stargate, Reaper, etc.)

Inspecting the cluster status

The cluster status can be obtained with the following command (executed in the appropriate namespace):

kubectl describe k8c <cluster_name>

Overall status

A K8ssandraCluster status has the following overall structure:

Status:
  Conditions: ...         # conditions applying to the whole cluster – see below
  Decommission Progress:  # decommission progress, if a datacenter is being decommissioned – see below 
  Datacenters:            # status of each managed datacenter in this cluster, keyed by name
    <datacenter_name>:
      Cassandra: ...      # status of the datacenter itself (always present)
      Reaper: ...         # status of Reaper, if deployed in this datacenter, absent otherwise
      Stargate: ...       # status of Stargate, if deployed in this datacenter, absent otherwise

The Datacenters entry is a map keyed by datacenter name. Each datacenter reports its own status per component: currently Cassandra, Reaper and Stargate statuses are included.

CassandraDatacenter status

The Cassandra entry of a datacenter status section is provided by cass-operator. The contents of this entry correspond to the status of the CassandraDatacenter resource, and provide useful information about the Cassandra cluster and its nodes.

When the datacenter is ready, the status of this entry looks like below:

# Status.Datacenters.<datacenter_name>:
  Cassandra:
    Cassandra Operator Progress:  Ready
    Conditions: ...
    Node Statuses:
      <pod_name>:
      <pod_name>:
      ...
    Observed Generation:  1
    Quiet Period:         2022-02-28T17:14:16Z
    Super User Upserted:  2022-02-28T17:14:11Z
    Users Upserted:       2022-02-28T17:14:11Z

Check cass-operator documentation for more information about the CassandraDatacenter resource status, and especially about all the conditions available, and their meanings.

Reaper status

The Reaper entry of a datacenter status is provided by k8ssandra-operator. The contents of this entry correspond to the status of the Reaper resource.

When Reaper is being deployed, this entry usually looks like below:

# Status.Datacenters.<datacenter_name>:
  Reaper:
    Conditions:
      Last Transition Time:  2022-02-28T17:20:04Z
      Status:                False
      Type:                  Ready
    Progress:                Configuring

Currently, Reaper only supports the Ready condition; it is set to true when Reaper is ready.

The Progress field can have the following values:

  • Pending: when the controller is waiting for the CassandraDatacenter to become ready.
  • Deploying: when controller is waiting for the Reaper deployment and its associated service to become ready.
  • Configuring: when the Reaper instance is ready for work and is being connected to its target datacenter.
  • Running: when Reaper is up and running.

When Reaper is ready, the status of this entry looks like below:

# Status.Datacenters.<datacenter_name>:
  Reaper:
    Conditions:
      Last Transition Time:  2022-02-28T17:22:35Z
      Status:                True
      Type:                  Ready
    Progress:                Running    

When Reaper is fully deployed, the Ready condition must be true, and the Progress field must be set to Running.

Stargate status

The Stargate entry of a datacenter status is provided by k8ssandra-operator. The contents of this entry correspond to the status of the Stargate resource.

When Stargate is being deployed, this entry usually looks like below:

# Status.Datacenters.<datacenter_name>:
  Stargate:
    Available Replicas:  0
    Conditions:
      Last Transition Time:  2022-02-28T17:22:42Z
      Status:                False
      Type:                  Ready
    Deployment Refs:
      <stargate_deployment_ref>
      <stargate_deployment_ref>
      ...
    Progress:              Deploying
    Ready Replicas:        0
    Ready Replicas Ratio:  0/3
    Replicas:              3
    Updated Replicas:      3

Currently, Stargate only supports the Ready condition; it is set to true when Stargate is ready.

The Progress field can have the following values:

  • Pending: when the controller is waiting for the datacenter to become ready.
  • Deploying: when the controller is waiting for the Stargate deployment and its associated service to become ready.
  • Running: when Stargate is up and running.

When Stargate is ready, the status of this entry looks like below:

# Status.Datacenters.<datacenter_name>:
  Stargate:
    Available Replicas:  3
    Conditions:
      Last Transition Time:  2022-02-28T17:20:01Z
      Status:                True
      Type:                  Ready
    Deployment Refs:
      <stargate_deployment_ref>
      <stargate_deployment_ref>
      ...
    Progress:              Running
    Ready Replicas:        3
    Ready Replicas Ratio:  3/3
    Replicas:              3
    Service Ref:           <service_ref>
    Updated Replicas:      3

When Stargate is fully deployed, the Ready condition must be true, and the Progress field must be set to Running.

Available K8ssandraCluster conditions

Currently, the only condition supported at K8ssandraCluster level is CassandraInitialized: it is set to true when the Cassandra cluster (that is, the Cassandra nodes without taking into account other components, such as Stargate or Reaper) becomes ready for the first time. During the lifetime of that Cassandra cluster, datacenters may have their readiness condition change back and forth. Once set, this condition however does not change. This condition is mainly intended for internal use.

Decommission Progress

The field Decommission Progress is only set when there is an ongoing datacenter decommission. When non-empty, it can have the following values:

  • UpdatingReplication: in this phase, keyspace replications are being updated to reflect the datacenter decommission.
  • Decommissioning: this phase is carried out by cass-operator and corresponds to the actual datacenter decommission.

Check quotas

In some cases, pods can become “unhealthy” and the root cause may be an insufficient quota. You can check quotas in the cloud provider’s UI. For example, in the Google Cloud Platform (GCP) console, check for any unhealthy pods in your GKE project. Then in the IAM & Admin section of the GCP console, navigate to Quotas. Look for any reported issues with backend services:

Backend service quota error

From the GCP > IAM & Admin > Quotas display:

  1. Select the row for the service name that is reporting a quota issue
  2. Click All Quotas from the Details column
  3. Check the box for the affected quota, and click Edit Quota.
  4. The dialog indicates: “Enter a new quota limit. Your request will be sent to your service provider for approval.” Examine the displayed current value and set a new value.
  5. Enter a brief request description and click Next.
  6. Verify your contact information, and click Submit Request.

Notice how in the following example the Backend services quota is set to ‘5’, and we’re changing it to ‘50’. For the K8ssandra deployments (Stargate, cass-operator, Reaper, Medusa, and so on), actually 10 might be a sufficient quota.

Quota UI showing change in Backend service quota from 5 to 50

Bucket region or name for backups is misconfigured

Among the operators installed by K8ssandra is Medusa, which provides backup and restore for Cassandra data.

If the storage object’s name or region used by an Amazon S3 bucket does not match the values expected by Medusa, an error is written to the Medusa section of the logs. Example:

kubectl logs demo-dc1-default-sts-0 -c medusa
.
.
.
File "/usr/local/lib/python3.6/dist-packages/libcloud/storage/drivers/s3.py", line 143, in parse_error driver=S3StorageDriver)
libcloud.common.types.LibcloudError: <LibcloudError in <class 'libcloud.storage.drivers.s3.S3StorageDriver'> 
'This bucket is located in a different region. Please use the correct driver. Bucket region "us-east-2", used region "us-east-1".'>
As a result of the region mismatch, the Medusa container within the <cluster-name>-dc1-default-sts-0 pod fails to start. While other pods launched by the K8ssandra install may start successfully, the <cluster-name>-dc1-default-sts-0 pod will not due to the Medusa error.

Separately in Amazon AWS, confirm that you know the correct region and name to use for your bucket. Example:

Amazon S3 Bucket Overview shows the name of the region

Declare the appropriate name and region in a values YAML. For example, create a file named my-backup-restore-values.yaml. Notice below the storage_properties setting for the region us-east-1, which matches the region configured and shown in the Amazon S3 user interface:

size: 3
backupRestore: 
  medusa:
    enabled: true
    bucketName: jsmart-k8ssandra-bucket2
    bucketSecret: medusa-bucket-key
    storage: s3
    storage_properties:
      region: us-east-1

Also make sure the bucketName matches: jsmart-k8ssandra-bucket2, in this example.

For example, relreferring again to the S3 UI, confirm the bucket name:

Confirm the bucket name as shown in the Amazon S3 UI

Then for a new or existing K8ssandra installation, reference the values file.

New install:

helm install demo k8ssandra/k8ssandra -f my-backup-restore-values.yaml

Upgrade:

helm upgrade demo k8ssandra/k8ssandra -f my-backup-restore-values.yaml

Incorrect credentials are configured for backups

If the Medusa log reports an authentication error, check that you provided the correct credentials. For example, with Amazon S3 buckets, check the credentials in the configured aws_access_key_id and aws_secret_access_key settings.

For example, my-medusa-bucket-key.yaml contains:

apiVersion: v1
kind: Secret
metadata:
 name: medusa-bucket-key
type: Opaque
stringData:
# Note that this currently has to be set to medusa_s3_credentials!
medusa_s3_credentials: |-
  [default]
  aws_access_key_id = FakeValues99ESPW3ALMEZ6U
  aws_secret_access_key = FakeValues99cl9bqJFVA3iFUm+yqVe08HxhXFE/ilK  

If your IT group manages S3 credentials, contact IT to get the correct values.

Before installing or upgrading K8ssandra, and before starting a backup, apply the Medusa bucket values to your Kubernetes environment. Example:

kubectl apply -f my-medusa-bucket-key.yaml

Output:

 secret/medusa-bucket-key configured

Collect useful information

Suppose you have an error after editing a K8ssandra configuration, or you want to inspect some things as you learn. There are some useful commands that come in handy when needing to dig a bit deeper. The following examples assume you are using a k8ssandra namespace, but this can be adjusted as needed.

Issue the following kubectl command to view the Management-api logs. Replace cassandra-pod with an actual pod instance name:

kubectl logs *cassandra-pod* -c cassandra -n k8ssandra

Issue the following kubectl command to view the Cassandra logs. Replace cassandra-pod with an actual pod instance name:

kubectl logs *cassandra-pod* -c server-system-logger -n k8ssandra

Issue the following kubectl command to view Medusa logs. Replace cassandra-pod with an actual pod instance name:

kubectl logs *cassandra-pod* -c medusa -n k8ssandra

Issue the following kubectl command to describe the CassandraDatacenter resource. This provides a wealth of information about the resource, which includes aged events that assist when trying to troubleshoot an issue:

kubectl describe cassandradatacenter/dc1 -n k8ssandra

Gather container specific information for a pod.

First, list out the pods scoped to the K8ssandra namespace and instance with a target release:

kubectl get pods -l app.kubernetes.io/instance=*release-name* -n k8ssandra

Next, targeting a specific pod, filter out container specific information. Replace the name of the pod with the pod of interest:

kubectl describe pod/*pod-name* -n k8ssandra | grep container -C 3

A slight variation: list out pods having the label for a cassandra cluster:

kubectl get pods -l cassandra.datastax.com/cluster=*release-name* -n k8ssandra

Now, using a pod-name returned, describe all the details:

kubectl describe pod/*pod-name* -n k8ssandra

Next steps

  • Explore other K8ssandra Operator tasks.
  • See the Reference topics for information about K8ssandra Operator Custom Resource Definitions (CRDs) and the single K8ssandra Operator Helm chart.


Last modified December 8, 2022: Fix installation docs (#1546) (9184626)