Deploy TiDB on AWS EKS

This document describes how to deploy a TiDB cluster on AWS Elastic Kubernetes Service (EKS).

To deploy TiDB Operator and the TiDB cluster in a self-managed Kubernetes environment, refer to Deploy TiDB Operator and Deploy TiDB on General Kubernetes.

Prerequisites

Before deploying a TiDB cluster on AWS EKS, make sure the following requirements are satisfied:

Install Helm 3: used for deploying TiDB Operator.
Complete all operations in Getting started with eksctl.
This guide includes the following contents:
- Install and configure awscli.
- Install and configure eksctl used for creating Kubernetes clusters.
- Install kubectl.

To verify whether AWS CLI is configured correctly, run the aws configure list command. If the output shows the values for access_key and secret_key, AWS CLI is configured correctly. Otherwise, you need to re-configure AWS CLI.

Note

The operations described in this document require at least the minimum privileges needed by eksctl and the service privileges needed to create a Linux bastion host.

Recommended instance types and storage

Instance types: to gain better performance, the following is recommended:
- PD nodes: c5.xlarge
- TiDB nodes: c5.4xlarge
- TiKV or TiFlash nodes: m5.4xlarge
Storage: Because AWS supports the EBS gp3 volume type, it is recommended to use EBS gp3. For gp3 provisioning, the following is recommended:
- TiKV: 400 MiB/s, 4000 IOPS
- TiFlash: 625 MiB/s, 6000 IOPS
AMI type: Amazon Linux 2

Create an EKS cluster and a node pool

According to AWS Official Blog recommendation and EKS Best Practice Document, since most of the TiDB cluster components use EBS volumes as storage, it is recommended to create a node pool in each availability zone (at least 3 in total) for each component when creating an EKS.

Save the following configuration as the cluster.yaml file. Replace ${clusterName} with your desired cluster name. The cluster and node group names should match the regular expression [a-zA-Z][-a-zA-Z0-9]*, so avoid names that contain _.

apiVersion: eksctl.io/v1alpha5
kind: ClusterConfig
metadata:
  name: ${clusterName}
  region: ap-northeast-1

nodeGroups:
  - name: admin
    desiredCapacity: 1
    privateNetworking: true
    labels:
      dedicated: admin

  - name: tidb-1a
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1a"]
    instanceType: c5.2xlarge
    labels:
      dedicated: tidb
    taints:
      dedicated: tidb:NoSchedule
  - name: tidb-1d
    desiredCapacity: 0
    privateNetworking: true
    availabilityZones: ["ap-northeast-1d"]
    instanceType: c5.2xlarge
    labels:
      dedicated: tidb
    taints:
      dedicated: tidb:NoSchedule
  - name: tidb-1c
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1c"]
    instanceType: c5.2xlarge
    labels:
      dedicated: tidb
    taints:
      dedicated: tidb:NoSchedule

  - name: pd-1a
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1a"]
    instanceType: c5.xlarge
    labels:
      dedicated: pd
    taints:
      dedicated: pd:NoSchedule
  - name: pd-1d
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1d"]
    instanceType: c5.xlarge
    labels:
      dedicated: pd
    taints:
      dedicated: pd:NoSchedule
  - name: pd-1c
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1c"]
    instanceType: c5.xlarge
    labels:
      dedicated: pd
    taints:
      dedicated: pd:NoSchedule

  - name: tikv-1a
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1a"]
    instanceType: r5b.2xlarge
    labels:
      dedicated: tikv
    taints:
      dedicated: tikv:NoSchedule
  - name: tikv-1d
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1d"]
    instanceType: r5b.2xlarge
    labels:
      dedicated: tikv
    taints:
      dedicated: tikv:NoSchedule
  - name: tikv-1c
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1c"]
    instanceType: r5b.2xlarge
    labels:
      dedicated: tikv
    taints:
      dedicated: tikv:NoSchedule

By default, only two TiDB nodes are required, so you can set the desiredCapacity of the tidb-1d node group to 0. You can scale out this node group any time if necessary.

Execute the following command to create the cluster:

eksctl create cluster -f cluster.yaml

After executing the command above, you need to wait until the EKS cluster is successfully created and the node group is created and added in the EKS cluster. This process might take 5 to 20 minutes. For more cluster configuration, refer to eksctl documentation.

Warning

If the Regional Auto Scaling Group (ASG) is used:

Enable the instance scale-in protection for all the EC2s that have been started. The instance scale-in protection for the ASG is not required.
Set termination policy to NewestInstance for the ASG.

Configure StorageClass

This section describes how to configure the storage class for different storage types. These storage types are:

The default gp2 storage type after creating the EKS cluster.
The gp3 storage type (recommended) or other EBS storage types.
The local storage used for testing bare-metal performance.

Configure `gp2`

note:
Starting from EKS Kubernetes 1.23, you need to deploy the EBS CSI driver before using the default gp2 storage class. For details, refer to the notice for Amazon EKS Kubernetes 1.23.

After you create an EKS cluster, the default StorageClass is gp2. To improve I/O write performance, it is recommended to configure nodelalloc and noatime in the mountOptions field of the StorageClass resource.

kind: StorageClass
apiVersion: storage.k8s.io/v1
# ...
mountOptions:
- nodelalloc,noatime

For more information on the mount options, see TiDB Environment and System Configuration Check.

Configure `gp3` (recommended) or other EBS storage types

If you do not want to use the default gp2 storage type, you can create StorageClass for other storage types. For example, you can use the gp3 (recommended) or io1 storage type.

The following example shows how to create and configure a StorageClass for the gp3 storage type:

Deploy the AWS EBS Container Storage Interface (CSI) driver on the EKS cluster. If you are using a storage type other than gp3, skip this step.

Set ebs-csi-node toleration.

kubectl patch -n kube-system ds ebs-csi-node -p '{"spec":{"template":{"spec":{"tolerations":[{"operator":"Exists"}]}}}}'

Expected output:

daemonset.apps/ebs-csi-node patched

Create a StorageClass resource. In the resource definition, specify your desired storage type in the parameters.type field.

kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
  name: gp3
provisioner: ebs.csi.aws.com
allowVolumeExpansion: true
volumeBindingMode: WaitForFirstConsumer
parameters:
  type: gp3
  fsType: ext4
  iops: "4000"
  throughput: "400"
mountOptions:
- nodelalloc,noatime

In the TidbCluster YAML file, configure gp3 in the storageClassName field. For example:
```
spec:
  tikv:
    ...
    storageClassName: gp3
```
To improve I/O write performance, it is recommended to configure nodelalloc and noatime in the mountOptions field of the StorageClass resource.
```
kind: StorageClass
apiVersion: storage.k8s.io/v1
# ...
mountOptions:
- nodelalloc,noatime
```
For more information on the mount options, see TiDB Environment and System Configuration Check.

For more information on the EBS storage types and configuration, refer to Amazon EBS volume types and Storage Classes.

Configure local storage

Local storage is used for testing bare-metal performance. For higher IOPS and lower latency, you can choose NVMe SSD volumes offered by some AWS instances for the TiKV node pool. However, for the production environment, use AWS EBS as your storage type.

Note

You cannot dynamically change StorageClass for a running TiDB cluster. For testing purposes, create a new TiDB cluster with the desired StorageClass.
EKS upgrade or other reasons might cause node reconstruction. In such cases, data in the local storage might be lost. To avoid data loss, you need to back up TiKV data before node reconstruction.
To avoid data loss from node reconstruction, you can refer to AWS documentation and disable the ReplaceUnhealthy feature of the TiKV node group.

For instance types that provide NVMe SSD volumes, check out Amazon EC2 Instance Types.

The following c5d.4xlarge example shows how to configure StorageClass for the local storage:

Create a node group with local storage for TiKV.
1. In the eksctl configuration file, modify the instance type of the TiKV node group to c5d.4xlarge:
```
  - name: tikv-1a
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1a"]
    instanceType: c5d.4xlarge
    labels:
      dedicated: tikv
    taints:
      dedicated: tikv:NoSchedule
    ...
```
2. Create a node group with local storage:
```
eksctl create nodegroups -f cluster.yaml
```
If the TiKV node group already exists, to avoid name conflict, you can take either of the following actions:
- Delete the old group and create a new one.
- Change the group name.
Deploy local volume provisioner.
1. To conveniently discover and manage local storage volumes, install local-volume-provisioner.
2. Mount the local storage to the /mnt/ssd directory.
3. According to the mounting configuration, modify the local-volume-provisioner.yaml file.
4. Deploy and create a local-storage storage class using the modified local-volume-provisioner.yaml file.
```
kubectl apply -f <local-volume-provisioner.yaml>
```
Use the local storage.
After you complete the previous step, local-volume-provisioner can discover all the local NVMe SSD volumes in the cluster.

After local-volume-provisioner discovers the local volumes, when you Deploy a TiDB cluster and the monitoring component, you need to add the tikv.storageClassName field to tidb-cluster.yaml and set the field value to local-storage.

Deploy TiDB Operator

To deploy TiDB Operator in the EKS cluster, refer to the Deploy TiDB Operator section in Getting Started.

Deploy a TiDB cluster and the monitoring component

This section describes how to deploy a TiDB cluster and its monitoring component in AWS EKS.

Create namespace

To create a namespace to deploy the TiDB cluster, run the following command:

kubectl create namespace tidb-cluster

Note

A namespace is a virtual cluster backed by the same physical cluster. This document takes tidb-cluster as an example. If you want to use another namespace, modify the corresponding arguments of -n or --namespace.

Deploy

First, download the sample TidbCluster and TidbMonitor configuration files:

curl -O https://raw.githubusercontent.com/pingcap/tidb-operator/v1.4.7/examples/aws/tidb-cluster.yaml && \
curl -O https://raw.githubusercontent.com/pingcap/tidb-operator/v1.4.7/examples/aws/tidb-monitor.yaml && \
curl -O https://raw.githubusercontent.com/pingcap/tidb-operator/v1.4.7/examples/aws/tidb-dashboard.yaml

Refer to configure the TiDB cluster to further customize and configure the CR before applying.

Note

By default, the configuration in tidb-cluster.yaml sets up the LoadBalancer for TiDB with the "internal" scheme. This means that the LoadBalancer is only accessible within the VPC, not externally. To access TiDB over the MySQL protocol, you need to use a bastion host or use kubectl port-forward. If you want to expose TiDB over the internet and if you are aware of the risks of doing this, you can change the scheme for the LoadBalancer from "internal" to "internet-facing" in the tidb-cluster.yaml file.

To deploy the TidbCluster and TidbMonitor CR in the EKS cluster, run the following command:

kubectl apply -f tidb-cluster.yaml -n tidb-cluster && \
kubectl apply -f tidb-monitor.yaml -n tidb-cluster

After the YAML file above is applied to the Kubernetes cluster, TiDB Operator creates the desired TiDB cluster and its monitoring component according to the YAML file.

Note

If you need to deploy a TiDB cluster on ARM64 machines, refer to Deploy a TiDB Cluster on ARM64 Machines.

View the cluster status

To view the status of the starting TiDB cluster, run the following command:

kubectl get pods -n tidb-cluster

When all the Pods are in the Running or Ready state, the TiDB cluster is successfully started. For example:

NAME                              READY   STATUS    RESTARTS   AGE
tidb-discovery-5cb8474d89-n8cxk   1/1     Running   0          47h
tidb-monitor-6fbcc68669-dsjlc     3/3     Running   0          47h
tidb-pd-0                         1/1     Running   0          47h
tidb-pd-1                         1/1     Running   0          46h
tidb-pd-2                         1/1     Running   0          46h
tidb-tidb-0                       2/2     Running   0          47h
tidb-tidb-1                       2/2     Running   0          46h
tidb-tikv-0                       1/1     Running   0          47h
tidb-tikv-1                       1/1     Running   0          47h
tidb-tikv-2                       1/1     Running   0          47h

Access the database

After you have deployed a TiDB cluster, you can access the TiDB database to test or develop your application.

Prepare a bastion host

The LoadBalancer created for your TiDB cluster is an intranet LoadBalancer. You can create a bastion host in the cluster VPC to access the database. To create a bastion host on AWS console, refer to AWS documentation.

Select the cluster's VPC and Subnet, and verify whether the cluster name is correct in the dropdown box. You can view the cluster's VPC and Subnet by running the following command:

eksctl get cluster -n ${clusterName}

Allow the bastion host to access the Internet. Select the correct key pair so that you can log in to the host via SSH.

Note

In addition to the bastion host, you can also connect an existing host to the cluster VPC by VPC Peering. If the EKS cluster is created in an existing VPC, you can use the host in the VPC.

Install the MySQL client and connect

After the bastion host is created, you can connect to the bastion host via SSH and access the TiDB cluster via the MySQL client.

ssh [-i /path/to/your/private-key.pem] ec2-user@<bastion-public-dns-name>

Install the MySQL client on the bastion host:
```
sudo yum install mysql -y
```

Connect the client to the TiDB cluster:

mysql --comments -h ${tidb-nlb-dnsname} -P 4000 -u root

${tidb-nlb-dnsname} is the LoadBalancer domain name of the TiDB service. You can view the domain name in the EXTERNAL-IP field by executing kubectl get svc basic-tidb -n tidb-cluster.

For example:

$ mysql --comments -h abfc623004ccb4cc3b363f3f37475af1-9774d22c27310bc1.elb.us-west-2.amazonaws.com -P 4000 -u root
Welcome to the MariaDB monitor.  Commands end with ; or \g.
Your MySQL connection id is 1189
Server version: 5.7.25-TiDB-v7.5.0 TiDB Server (Apache License 2.0) Community Edition, MySQL 5.7 compatible

Copyright (c) 2000, 2022, Oracle and/or its affiliates.

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

MySQL [(none)]> show status;
+--------------------+--------------------------------------+
| Variable_name      | Value                                |
+--------------------+--------------------------------------+
| Ssl_cipher         |                                      |
| Ssl_cipher_list    |                                      |
| Ssl_verify_mode    | 0                                    |
| Ssl_version        |                                      |
| ddl_schema_version | 22                                   |
| server_id          | ed4ba88b-436a-424d-9087-977e897cf5ec |
+--------------------+--------------------------------------+
6 rows in set (0.00 sec)

Note

The default authentication plugin of MySQL 8.0 is updated from mysql_native_password to caching_sha2_password. Therefore, if you use MySQL client from MySQL 8.0 to access the TiDB service (cluster version < v4.0.7), and if the user account has a password, you need to explicitly specify the --default-auth=mysql_native_password parameter.
By default, TiDB (starting from v4.0.2) periodically shares usage details with PingCAP to help understand how to improve the product. For details about what is shared and how to disable the sharing, see Telemetry.

Access the Grafana monitoring dashboard

Obtain the LoadBalancer domain name of Grafana:

kubectl -n tidb-cluster get svc basic-grafana

For example:

$ kubectl get svc basic-grafana
NAME            TYPE           CLUSTER-IP      EXTERNAL-IP                                                             PORT(S)          AGE
basic-grafana   LoadBalancer   10.100.199.42   a806cfe84c12a4831aa3313e792e3eed-1964630135.us-west-2.elb.amazonaws.com 3000:30761/TCP   121m

In the output above, the EXTERNAL-IP column is the LoadBalancer domain name.

You can access the ${grafana-lb}:3000 address using your web browser to view monitoring metrics. Replace ${grafana-lb} with the LoadBalancer domain name.

Note

The default Grafana username and password are both admin.

Access the TiDB Dashboard

See Access TiDB Dashboard for instructions about how to securely allow access to the TiDB Dashboard.

Upgrade

To upgrade the TiDB cluster, execute the following command:

kubectl patch tc basic -n tidb-cluster --type merge -p '{"spec":{"version":"${version}"}}`.

The upgrade process does not finish immediately. You can watch the upgrade progress by executing kubectl get pods -n tidb-cluster --watch.

Scale out

Before scaling out the cluster, you need to scale out the corresponding node group so that the new instances have enough resources for operation.

This section describes how to scale out the EKS node group and TiDB components.

Scale out EKS node group

When scaling out TiKV, the node groups must be scaled out evenly among the different availability zones. The following example shows how to scale out the tikv-1a, tikv-1c, and tikv-1d groups of the ${clusterName} cluster to 2 nodes:

eksctl scale nodegroup --cluster ${clusterName} --name tikv-1a --nodes 2 --nodes-min 2 --nodes-max 2
eksctl scale nodegroup --cluster ${clusterName} --name tikv-1c --nodes 2 --nodes-min 2 --nodes-max 2
eksctl scale nodegroup --cluster ${clusterName} --name tikv-1d --nodes 2 --nodes-min 2 --nodes-max 2

For more information on managing node groups, refer to eksctl documentation.

Scale out TiDB components

After scaling out the EKS node group, execute kubectl edit tc basic -n tidb-cluster, and modify each component's replicas to the desired number of replicas. The scaling-out process is then completed.

Deploy TiFlash/TiCDC

TiFlash is the columnar storage extension of TiKV.

TiCDC is a tool for replicating the incremental data of TiDB by pulling TiKV change logs.

The two components are not required in the deployment. This section shows a quick start example.

Add node groups

In the configuration file of eksctl (cluster.yaml), add the following two items to add a node group for TiFlash/TiCDC respectively. desiredCapacity is the number of nodes you desire.

  - name: tiflash-1a
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1a"]
    labels:
      dedicated: tiflash
    taints:
      dedicated: tiflash:NoSchedule
  - name: tiflash-1d
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1d"]
    labels:
      dedicated: tiflash
    taints:
      dedicated: tiflash:NoSchedule
  - name: tiflash-1c
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1c"]
    labels:
      dedicated: tiflash
    taints:
      dedicated: tiflash:NoSchedule

  - name: ticdc-1a
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1a"]
    labels:
      dedicated: ticdc
    taints:
      dedicated: ticdc:NoSchedule
  - name: ticdc-1d
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1d"]
    labels:
      dedicated: ticdc
    taints:
      dedicated: ticdc:NoSchedule
  - name: ticdc-1c
    desiredCapacity: 1
    privateNetworking: true
    availabilityZones: ["ap-northeast-1c"]
    labels:
      dedicated: ticdc
    taints:
      dedicated: ticdc:NoSchedule

Depending on the EKS cluster status, use different commands:

If the cluster is not created, execute eksctl create cluster -f cluster.yaml to create the cluster and node groups.
If the cluster is already created, execute eksctl create nodegroup -f cluster.yaml to create the node groups. The existing node groups are ignored and will not be created again.

Configure and deploy

To deploy TiFlash, configure spec.tiflash in tidb-cluster.yaml:
```
spec:
  ...
  tiflash:
    baseImage: pingcap/tiflash
    maxFailoverCount: 0
    replicas: 1
    storageClaims:
    - resources:
        requests:
          storage: 100Gi
    tolerations:
    - effect: NoSchedule
      key: dedicated
      operator: Equal
      value: tiflash
```
For other parameters, refer to Configure a TiDB Cluster.
Warning
TiDB Operator automatically mount PVs in the order of the configuration in the storageClaims list. Therefore, if you need to add disks for TiFlash, make sure that you add the disks only to the end of the original configuration in the list. In addition, you must not alter the order of the original configuration.

To deploy TiCDC, configure spec.ticdc in tidb-cluster.yaml:

spec:
  ...
  ticdc:
    baseImage: pingcap/ticdc
    replicas: 1
    tolerations:
    - effect: NoSchedule
      key: dedicated
      operator: Equal
      value: ticdc

Modify replicas according to your needs.

Finally, execute kubectl -n tidb-cluster apply -f tidb-cluster.yaml to update the TiDB cluster configuration.

For detailed CR configuration, refer to API references and Configure a TiDB Cluster.

Configure TiDB monitoring

For more information, see Deploy monitoring and alerts for a TiDB cluster.

Note

TiDB monitoring does not persist data by default. To ensure long-term data availability, it is recommended to persist monitoring data. TiDB monitoring does not include Pod CPU, memory, or disk monitoring, nor does it have an alerting system. For more comprehensive monitoring and alerting, it is recommended to Set kube-prometheus and AlertManager.

Collect logs

System and application logs can be useful for troubleshooting issues and automating operations. By default, TiDB components output logs to the container's stdout and stderr, and log rotation is automatically performed based on the container runtime environment. When a Pod restarts, container logs will be lost. To prevent log loss, it is recommended to Collect logs of TiDB and its related components.

Deploy TiDB on AWS EKS

Prerequisites

Recommended instance types and storage

Create an EKS cluster and a node pool

Configure StorageClass

Configure gp2

Configure gp3 (recommended) or other EBS storage types

Configure local storage

Deploy TiDB Operator

Deploy a TiDB cluster and the monitoring component

Create namespace

Deploy

View the cluster status

Access the database

Prepare a bastion host

Install the MySQL client and connect

Access the Grafana monitoring dashboard

Access the TiDB Dashboard

Upgrade

Scale out

Scale out EKS node group

Scale out TiDB components

Deploy TiFlash/TiCDC

Add node groups

Configure and deploy

Configure TiDB monitoring

Collect logs

Was this page helpful?

Configure `gp2`

Configure `gp3` (recommended) or other EBS storage types