Skip to content

Latest commit

 

History

History
150 lines (140 loc) · 7.96 KB

chi_update_add_replication.md

File metadata and controls

150 lines (140 loc) · 7.96 KB
Raw

Update ClickHouse Installation - add replication to existing non-replicated cluster

Prerequisites

  1. Assume we have clickhouse-operator already installed and running
  2. Assume we have Zookeeper already installed and running

Install ClickHouseInstallation example

We are going to install everything into dev namespace. clickhouse-operator is already installed into dev namespace

Check pre-initial position.

kubectl -n dev get all,configmap

clickhouse-operator-only is available

NAME                                      READY   STATUS    RESTARTS   AGE
pod/clickhouse-operator-5cbc47484-v5cfg   1/1     Running   0          17s
NAME                                  TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)    AGE
service/clickhouse-operator-metrics   ClusterIP   10.102.229.22   <none>        8888/TCP   17s
NAME                                  READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/clickhouse-operator   1/1     1            1           17s
NAME                                            DESIRED   CURRENT   READY   AGE
replicaset.apps/clickhouse-operator-5cbc47484   1         1         1       17s

Now let's install ClickHouse from provided examples. There are two rolling update examples presented:

  1. Simple stateless cluster: initial position and update
  2. Stateful cluster with Persistent Volumes: initial position and update

Simple Rolling Update Example

Let's go with simple stateless cluster. Manifest file with initial position:

kubectl -n dev apply -f 07-rolling-update-stateless-01-initial-position.yaml

Check initial position. We should have cluster up and running:

kubectl -n dev get all,configmap

ClickHouse is installed, up and running, all 4 expected pods are running

NAME                                      READY   STATUS    RESTARTS   AGE
pod/chi-d02eaa-347e-0-0-0                 1/1     Running   0          19s
pod/chi-d02eaa-347e-1-0-0                 1/1     Running   0          19s
pod/chi-d02eaa-347e-2-0-0                 1/1     Running   0          19s
pod/chi-d02eaa-347e-3-0-0                 1/1     Running   0          19s
NAME                                  TYPE           CLUSTER-IP      EXTERNAL-IP   PORT(S)                         AGE
service/chi-d02eaa-347e-0-0           ClusterIP      None            <none>        8123/TCP,9000/TCP,9009/TCP      19s
service/chi-d02eaa-347e-1-0           ClusterIP      None            <none>        8123/TCP,9000/TCP,9009/TCP      19s
service/chi-d02eaa-347e-2-0           ClusterIP      None            <none>        8123/TCP,9000/TCP,9009/TCP      19s
service/chi-d02eaa-347e-3-0           ClusterIP      None            <none>        8123/TCP,9000/TCP,9009/TCP      19s
service/update-setup-replication      LoadBalancer   10.106.139.11   <pending>     8123:31328/TCP,9000:32091/TCP   19s
NAME                                   READY   AGE
statefulset.apps/chi-d02eaa-347e-0-0   1/1     19s
statefulset.apps/chi-d02eaa-347e-1-0   1/1     19s
statefulset.apps/chi-d02eaa-347e-2-0   1/1     19s
statefulset.apps/chi-d02eaa-347e-3-0   1/1     19s
NAME                                         DATA   AGE
configmap/chi-d02eaa-common-configd          2      19s
configmap/chi-d02eaa-common-usersd           0      19s
configmap/chi-d02eaa-deploy-confd-347e-0-0   1      19s
configmap/chi-d02eaa-deploy-confd-347e-1-0   1      19s
configmap/chi-d02eaa-deploy-confd-347e-2-0   1      19s
configmap/chi-d02eaa-deploy-confd-347e-3-0   1      19s

Let's explore one Pod in order to check available ClickHouse config files. Navigate directly inside the Pod:

kubectl -n dev exec -it chi-d02eaa-347e-0-0-0 -- bash

And take a look on ClickHouse config files available

root@chi-d02eaa-347e-0-0-0:/etc/clickhouse-server# ls /etc/clickhouse-server/conf.d/
macros.xml
root@chi-d02eaa-347e-0-0-0:/etc/clickhouse-server# ls /etc/clickhouse-server/config.d
listen.xml  remote_servers.xml

Standard minimal set of config files is in place. All is well.

Update ClickHouse Installation - add replication

Let's run update and change .yaml manifest so we'll have replication available.

In order to have replication correctly setup, we need to specify Zookeeper (which is assumed to be running already) and specify replicas for ClickHouse. Manifest file with updates specified:

kubectl -n dev apply -f 07-rolling-update-stateless-02-apply-update.yaml

And let's watch on how update is rolling over:

NAME                                      READY   STATUS              RESTARTS   AGE
pod/chi-d02eaa-347e-0-0-0                 0/1     ContainerCreating   0          1s
pod/chi-d02eaa-347e-1-0-0                 1/1     Running             0          4m41s
pod/chi-d02eaa-347e-2-0-0                 1/1     Running             0          4m41s
pod/chi-d02eaa-347e-3-0-0                 1/1     Running             0          4m41s

As we can see, the first Pod is being updated - pod/chi-d02eaa-347e-0-0-0. Then we can see how the first replica is being created - pod/chi-d02eaa-347e-0-1-0, as well as next Pod is being updated - pod/chi-d02eaa-347e-1-0-0

NAME                                      READY   STATUS              RESTARTS   AGE
pod/chi-d02eaa-347e-0-0-0                 1/1     Running             0          21s
pod/chi-d02eaa-347e-0-1-0                 0/1     ContainerCreating   0          2s
pod/chi-d02eaa-347e-1-0-0                 0/1     ContainerCreating   0          2s
pod/chi-d02eaa-347e-2-0-0                 1/1     Running             0          5m1s
pod/chi-d02eaa-347e-3-0-0                 1/1     Running             0          5m1s

Update is rolling further.

NAME                                      READY   STATUS              RESTARTS   AGE
pod/chi-d02eaa-347e-0-0-0                 1/1     Running             0          62s
pod/chi-d02eaa-347e-0-1-0                 1/1     Running             0          43s
pod/chi-d02eaa-347e-1-0-0                 1/1     Running             0          43s
pod/chi-d02eaa-347e-1-1-0                 1/1     Running             0          20s
pod/chi-d02eaa-347e-2-0-0                 1/1     Running             0          20s
pod/chi-d02eaa-347e-2-1-0                 0/1     ContainerCreating   0          1s
pod/chi-d02eaa-347e-3-0-0                 0/1     ContainerCreating   0          1s

And, after all Pod updated, let's take a look on ClickHouse config files available:

root@chi-d02eaa-347e-0-0-0:/# ls /etc/clickhouse-server/conf.d/
macros.xml
root@chi-d02eaa-347e-0-0-0:/# ls /etc/clickhouse-server/config.d
listen.xml  remote_servers.xml  zookeeper.xml

We can see new file added: zookeeper.xml. It is Zookeeper configuration for ClickHouse. Let's take a look into:

root@chi-d02eaa-347e-0-0-0:/# cat /etc/clickhouse-server/config.d/zookeeper.xml 
<yandex>
    <zookeeper>
        <node>
            <host>zookeeper.zoo1ns</host>
            <port>2181</port>
        </node>
    </zookeeper>
    <distributed_ddl>
        <path>/clickhouse/update-setup-replication/task_queue/ddl</path>
    </distributed_ddl>
</yandex>

Rolling Update with State Example

Stateful cluster with Persistent Volumes examples are presented as initial position and update The structure of the example is the same as for simple example, but Persistent Volumes are used. So this example is better to be run on cloud provider with Dynamic Volumes Provisioning available.