deprecated-helm-charts/stable/cockroachdb

CockroachDB Helm Chart

Prerequisites Details

• Kubernetes 1.8
• PV support on the underlying infrastructure. Docker for windows hostpath provisioner is not supported.
• If you want to secure your cluster to use TLS certificates for all network communication, Helm must be installed with RBAC privileges or else you will get an "attempt to grant extra privileges" error.

StatefulSet Details

• http://kubernetes.io/docs/concepts/abstractions/controllers/statefulsets/

StatefulSet Caveats

• http://kubernetes.io/docs/concepts/abstractions/controllers/statefulsets/#limitations

Chart Details

This chart will do the following:

• Set up a dynamically scalable CockroachDB cluster using a Kubernetes StatefulSet

Installing the Chart

To install the chart with the release name my-release:

helm install --name my-release stable/cockroachdb

Note that for a production cluster, you are very likely to want to modify the Storage and StorageClass parameters. This chart defaults to 100 GiB of disk space per pod, but you may want more or less depending on your use case, and the default persistent volume StorageClass in your environment may not be what you want for a database (e.g. on GCE and Azure the default is not SSD).

If you are running in secure mode (with configuration parameter Secure.Enabled set to true), then you will have to manually approve the cluster's security certificates as the pods are created. You can see the pending certificate-signing requests by running kubectl get csr, and approve them by running kubectl certificate approve <csr-name>. You'll have to approve one certificate for each node (e.g. default.node.eerie-horse-cockroachdb-0 and one client certificate for the job that initializes the cluster (e.g. default.node.root).

Upgrading

To 2.0.0

Due to having no explicit selector set for the StatefulSet before version 2.0.0 of this chart, upgrading from any version that uses a version of kubernetes that locks the selector labels to any other version is impossible without deleting the StatefulSet. Luckily there is a way to do it without actually deleting all the resources managed by the StatefulSet. Use the workaround below to upgrade from charts versions previous to 2.0.0. The following example assumes that the release name is crdb:

$ kubectl delete statefulset crdb-cockroachdb --cascade=false

Verify that no pod is deleted and then upgrade as normal. A new StatefulSet will be created taking over the management of the existing pods upgrading them if needed.

For more information about the upgrading bug see https://github.com/helm/charts/issues/7680.

Configuration

The following table lists the configurable parameters of the CockroachDB chart and their default values.

Parameter	Description	Default
Name	Chart name	cockroachdb
Image	Container image name	cockroachdb/cockroach
ImageTag	Container image tag	v2.1.5
ImagePullPolicy	Container pull policy	Always
Replicas	k8s statefulset replicas	3
MaxUnavailable	k8s PodDisruptionBudget parameter	1
Component	k8s selector key	cockroachdb
ExternalGrpcPort	CockroachDB primary serving port	26257
ExternalGrpcName	CockroachDB primary serving port name	grpc
InternalGrpcPort	CockroachDB inter-cockroachdb port	26257
InternalGrpcName	CockroachDB inter-cockroachdb port name	grpc
InternalHttpPort	CockroachDB HTTP port	8080
ExternalHttpPort	CockroachDB HTTP port on service	8080
HttpName	Name given to the http service port	http
Resources	Resource requests and limits	{}
Storage	Persistent volume size	100Gi
StorageClass	Persistent volume class	null
CacheSize	Size of CockroachDB's in-memory cache	25%
MaxSQLMemory	Max memory to use processing SQL queries	25%
ClusterDomain	Cluster's default DNS domain	cluster.local
NetworkPolicy.Enabled	Enable NetworkPolicy	false
NetworkPolicy.AllowExternal	Don't require client label for connections	true
Service.Type	Public service type	ClusterIP
Service.Annotations	Annotations to apply to the service	{}
PodManagementPolicy	OrderedReady or Parallel pod creation/deletion order	Parallel
UpdateStrategy.type	allows setting of RollingUpdate strategy	RollingUpdate
NodeSelector	Node labels for pod assignment	{}
Tolerations	List of node taints to tolerate	{}
Secure.Enabled	Whether to run securely using TLS certificates	false
Secure.RequestCertsImage	Image to use for requesting TLS certificates	cockroachdb/cockroach-k8s-request-cert
Secure.RequestCertsImageTag	Image tag to use for requesting TLS certificates	0.4
Secure.ServiceAccount.Create	Whether to create a new RBAC service account	true
Secure.ServiceAccount.Name	Name of RBAC service account to use	""
JoinExisting	List of already-existing cockroach instances	[]
Locality	Locality attribute for this deployment	""
ExtraArgs	Additional command-line arguments	[]
ExtraSecretMounts	Additional secrets to mount at cluster members	[]
ExtraEnvArgs	Allows to set extra ENV args	[]

Specify each parameter using the --set key=value[,key=value] argument to helm install.

Alternatively, a YAML file that specifies the values for the parameters can be provided while installing the chart. For example,

helm install --name my-release -f values.yaml stable/cockroachdb

Tip

: You can use the default values.yaml

Deep dive

Connecting to the CockroachDB cluster

Once you've created the cluster, you can start talking to it by connecting to its "public" service. CockroachDB is PostgreSQL wire protocol compatible so there's a wide variety of supported clients. For the sake of example, we'll open up a SQL shell using CockroachDB's built-in shell and play around with it a bit, like this (likely needing to replace "my-release-cockroachdb-public" with the name of the "-public" service that was created with your installed chart):

$ kubectl run -it --rm cockroach-client \
    --image=cockroachdb/cockroach \
    --restart=Never \
    --command -- ./cockroach sql --insecure --host my-release-cockroachdb-public
Waiting for pod default/cockroach-client to be running, status is Pending,
pod ready: false
If you don't see a command prompt, try pressing enter.
root@my-release-cockroachdb-public:26257> SHOW DATABASES;
+--------------------+
|      Database      |
+--------------------+
| information_schema |
| pg_catalog         |
| system             |
+--------------------+
(3 rows)
root@my-release-cockroachdb-public:26257> CREATE DATABASE bank;
CREATE DATABASE
root@my-release-cockroachdb-public:26257> CREATE TABLE bank.accounts (id INT
PRIMARY KEY, balance DECIMAL);
CREATE TABLE
root@my-release-cockroachdb-public:26257> INSERT INTO bank.accounts VALUES
(1234, 10000.50);
INSERT 1
root@my-release-cockroachdb-public:26257> SELECT * FROM bank.accounts;
+------+---------+
|  id  | balance |
+------+---------+
| 1234 | 10000.5 |
+------+---------+
(1 row)
root@my-release-cockroachdb-public:26257> \q
Waiting for pod default/cockroach-client to terminate, status is Running
pod "cockroach-client" deleted

If you are running in secure mode, you will have to provide a client certificate to the cluster in order to authenticate, so the above command will not work. See here for an example of how to set up an interactive SQL shell against a secure cluster or here for an example application connecting to a secure cluster.

Cluster health

Because our pod spec includes regular health checks of the CockroachDB processes, simply running kubectl get pods and looking at the STATUS column is sufficient to determine the health of each instance in the cluster.

If you want more detailed information about the cluster, the best place to look is the admin UI.

Accessing the admin UI

If you want to see information about how the cluster is doing, you can try pulling up the CockroachDB admin UI by port-forwarding from your local machine to one of the pods (replacing "my-release-cockroachdb-0" with one of your pods' names):

kubectl port-forward my-release-cockroachdb-0 8080

Once you’ve done that, you should be able to access the admin UI by visiting http://localhost:8080/ in your web browser.

Failover

If any CockroachDB member fails it gets restarted or recreated automatically by the Kubernetes infrastructure, and will rejoin the cluster automatically when it comes back up. You can test this scenario by killing any of the pods:

kubectl delete pod my-release-cockroachdb-1

$ kubectl get pods -l "component=my-release-cockroachdb"
NAME                      READY     STATUS        RESTARTS   AGE
my-release-cockroachdb-0  1/1       Running       0          5m
my-release-cockroachdb-2  1/1       Running       0          5m

After a while:

$ kubectl get pods -l "component=my-release-cockroachdb"
NAME                      READY     STATUS        RESTARTS   AGE
my-release-cockroachdb-0  1/1       Running       0          5m
my-release-cockroachdb-1  1/1       Running       0          20s
my-release-cockroachdb-2  1/1       Running       0          5m

You can check state of re-joining from the new pod's logs:

$ kubectl logs my-release-cockroachdb-1
[...]
I161028 19:32:09.754026 1 server/node.go:586  [n1] node connected via gossip and
verified as part of cluster {"35ecbc27-3f67-4e7d-9b8f-27c31aae17d6"}
[...]
cockroachdb-0.my-release-cockroachdb.default.svc.cluster.local:26257
build:      beta-20161027-55-gd2d3c7f @ 2016/10/28 19:27:25 (go1.7.3)
admin:      http://0.0.0.0:8080
sql:
postgresql://root@my-release-cockroachdb-1.my-release-cockroachdb.default.svc.cluster.local:26257?sslmode=disable
logs:       cockroach-data/logs
store[0]:   path=cockroach-data
status:     restarted pre-existing node
clusterID:  {35ecbc27-3f67-4e7d-9b8f-27c31aae17d6}
nodeID:     2
[...]

NetworkPolicy

To enable network policy for CockroachDB, install a networking plugin that implements the Kubernetes NetworkPolicy spec, and set NetworkPolicy.Enabled to true.

For Kubernetes v1.5 & v1.6, you must also turn on NetworkPolicy by setting the DefaultDeny namespace annotation. Note: this will enforce policy for all pods in the namespace:

kubectl annotate namespace default "net.beta.kubernetes.io/network-policy={\"ingress\":{\"isolation\":\"DefaultDeny\"}}"

For more precise policy, set networkPolicy.allowExternal=false. This will only allow pods with the generated client label to connect to CockroachDB. This label will be displayed in the output of a successful install.

Scaling

Scaling should typically be managed via the helm upgrade command, but StatefulSets don't yet work with helm upgrade. In the meantime until helm upgrade works, if you want to change the number of replicas, you can use the kubectl scale as shown below:

kubectl scale statefulset my-release-cockroachdb --replicas=4

Note that if you are running in secure mode and increase the size of your cluster, you will also have to approve the certificate-signing request of each new node (using kubectl get csr and kubectl certificate approve).