You are viewing the archived documentation of TiDB, which no longer receives updates.View latest LTS version docs ↗

Production Deployment from Binary Tarball

This guide provides installation instructions from a binary tarball on Linux. A complete TiDB cluster contains PD, TiKV, and TiDB. To start the database service, follow the order of PD -> TiKV -> TiDB. To stop the database service, follow the order of stopping TiDB -> TiKV -> PD.

See also local deployment and testing environment deployment.

Prepare

Before you start, see TiDB architecture and Software and Hardware Recommendations. Make sure the following requirements are satisfied:

Operating system

For the operating system, it is recommended to use RHEL/CentOS 7.3 or higher. The following additional requirements are recommended:

Configuration	Description
Supported Platform	RHEL/CentOS 7.3+ (more details)
File System	ext4 is recommended
Swap Space	Should be disabled
Disk Block Size	Set the system disk `Block` size to `4096`

Network and firewall

Configuration	Description
Firewall/Port	Check whether the ports required by TiDB are accessible between the nodes

Operating system parameters

Configuration	Description
Nice Limits	For system users, set the default value of `nice` in TiDB to `0`
min_free_kbytes	The setting for `vm.min_free_kbytes` in `sysctl.conf` needs to be high enough
User Open Files Limit	For database administrators, set the number of TiDB open files to `1000000`
System Open File Limits	Set the number of system open files to `1000000`
User Process Limits	For TiDB users, set the `nproc` value to `4096` in `limits.conf`
Address Space Limits	For TiDB users, set the space to `unlimited` in `limits.conf`
File Size Limits	For TiDB users, set the `fsize` value to `unlimited` in `limits.conf`
Disk Readahead	Set the value of the `readahead` data disk to `4096` at a minimum
NTP service	Configure the NTP time synchronization service for each node
SELinux	Turn off the SELinux service for each node
CPU Frequency Scaling	It is recommended to turn on CPU overclocking
Transparent Hugepages	For Red Hat 7+ and CentOS 7+ systems, it is required to set the Transparent Hugepages to `always`
I/O Scheduler	Set the I/O Scheduler of data disks to the `deadline` mode
vm.swappiness	Set `vm.swappiness = 0` in `sysctl.conf`
net.core.somaxconn	Set `net.core.somaxconn = 32768` in `sysctl.conf`
net.ipv4.tcp_syncookies	Set `net.ipv4.tcp_syncookies = 0` in `sysctl.conf`

Database running user settings

Configuration	Description
LANG environment	Set `LANG = en_US.UTF8`
TZ time zone	Set the TZ time zone of all nodes to the same value

TiDB components and default ports

Before you deploy a TiDB cluster, see the required components and optional components.

TiDB database components (required)

See the following table for the default ports for the TiDB components:

Component	Default Port	Protocol	Description
ssh	22	TCP	the sshd service
TiDB	4000	TCP	the communication port for the application and DBA tools
TiDB	10080	TCP	the communication port to report TiDB status
TiKV	20160	TCP	the TiKV communication port
PD	2379	TCP	the communication port between TiDB and PD
PD	2380	TCP	the inter-node communication port within the PD cluster

TiDB database components (optional)

See the following table for the default ports for the optional TiDB components:

Component	Default Port	Protocol	Description
Prometheus	9090	TCP	the communication port for the Prometheus service
Pushgateway	9091	TCP	the aggregation and report port for TiDB, TiKV, and PD monitor
Node_exporter	9100	TCP	the communication port to report the system information of every TiDB cluster node
Grafana	3000	TCP	the port for the external Web monitoring service and client (Browser) access
alertmanager	9093	TCP	the port for the alert service

Create a database running user account

Log in to the machine using the root user account and create a database running user account (tidb) using the following command:
```
useradd tidb -m
```
Switch the user from root to tidb by using the following command. You can use this tidb user account to deploy your TiDB cluster.
```
su - tidb
```

Download the official binary package

# Download the package.
$ wget https://download.pingcap.org/tidb-{version}-linux-amd64.tar.gz
$ wget https://download.pingcap.org/tidb-{version}-linux-amd64.sha256

# Check the file integrity. If the result is OK, the file is correct.
$ sha256sum -c tidb-{version}-linux-amd64.sha256

# Extract the package.
$ tar -xzf tidb-{version}-linux-amd64.tar.gz
$ cd tidb-{version}-linux-amd64

Note

{version} indicates the version number of TiDB. For example, if {version} is v3.0.0, the package download link is https://download.pingcap.org/tidb-v3.0.0-linux-amd64.tar.gz.

Multiple nodes cluster deployment

For the production environment, multiple nodes cluster deployment is recommended. Before you begin, see Software and Hardware Recommendations.

Assuming that you have six nodes, you can deploy 3 PD instances, 3 TiKV instances, and 1 TiDB instance. See the following table for details:

Name	Host IP	Services
Node1	192.168.199.113	PD1, TiDB
Node2	192.168.199.114	PD2
Node3	192.168.199.115	PD3
Node4	192.168.199.116	TiKV1
Node5	192.168.199.117	TiKV2
Node6	192.168.199.118	TiKV3

Follow the steps below to start PD, TiKV, and TiDB:

Start PD on Node1, Node2, and Node3 in sequence.

./bin/pd-server --name=pd1 \
                --data-dir=pd \
                --client-urls="http://192.168.199.113:2379" \
                --peer-urls="http://192.168.199.113:2380" \
                --initial-cluster="pd1=http://192.168.199.113:2380,pd2=http://192.168.199.114:2380,pd3=http://192.168.199.115:2380" \
                -L "info" \
                --log-file=pd.log &

./bin/pd-server --name=pd2 \
                --data-dir=pd \
                --client-urls="http://192.168.199.114:2379" \
                --peer-urls="http://192.168.199.114:2380" \
                --initial-cluster="pd1=http://192.168.199.113:2380,pd2=http://192.168.199.114:2380,pd3=http://192.168.199.115:2380" \
                -L "info" \
                --log-file=pd.log &

./bin/pd-server --name=pd3 \
                --data-dir=pd \
                --client-urls="http://192.168.199.115:2379" \
                --peer-urls="http://192.168.199.115:2380" \
                --initial-cluster="pd1=http://192.168.199.113:2380,pd2=http://192.168.199.114:2380,pd3=http://192.168.199.115:2380" \
                -L "info" \
                --log-file=pd.log &

Start TiKV on Node4, Node5 and Node6.

./bin/tikv-server --pd="192.168.199.113:2379,192.168.199.114:2379,192.168.199.115:2379" \
                  --addr="192.168.199.116:20160" \
                  --status-addr="192.168.199.116:20180" \
                  --data-dir=tikv \
                  --log-file=tikv.log &

./bin/tikv-server --pd="192.168.199.113:2379,192.168.199.114:2379,192.168.199.115:2379" \
                  --addr="192.168.199.117:20160" \
                  --status-addr="192.168.199.117:20180" \
                  --data-dir=tikv \
                  --log-file=tikv.log &

./bin/tikv-server --pd="192.168.199.113:2379,192.168.199.114:2379,192.168.199.115:2379" \
                  --addr="192.168.199.118:20160" \
                  --status-addr="192.168.199.118:20180" \
                  --data-dir=tikv \
                  --log-file=tikv.log &

Start TiDB on Node1.

./bin/tidb-server --store=tikv \
                  --path="192.168.199.113:2379,192.168.199.114:2379,192.168.199.115:2379" \
                  --log-file=tidb.log &

Use the MySQL client to connect to TiDB.

mysql -h 192.168.199.113 -P 4000 -u root -D test

Note

If you start TiKV or deploy PD in the production environment, it is highly recommended to specify the path for the configuration file using the --config parameter. If the parameter is not set, TiKV or PD does not read the configuration file.
To tune TiKV, see Performance Tuning for TiKV.
If you use nohup to start the cluster in the production environment, write the startup commands in a script and then run the script. If not, the nohup process might abort because it receives exceptions when the Shell command exits. For more information, see The TiDB/TiKV/PD process aborts unexpectedly.

For the deployment and use of TiDB monitoring services, see Monitor a TiDB Cluster.