title | summary | aliases | ||
---|---|---|---|---|
Software and Hardware Recommendations |
Learn the software and hardware recommendations for deploying and running TiDB. |
|
As an open-source distributed SQL database with high performance, TiDB can be deployed in the Intel architecture server, ARM architecture server, and major virtualization environments and runs well. TiDB supports most of the major hardware networks and Linux operating systems.
Operating systems | Supported CPU architectures |
---|---|
Red Hat Enterprise Linux 8.4 or a later 8.x version |
|
Red Hat Enterprise Linux 7.3 or a later 7.x version |
|
Amazon Linux 2 |
|
Amazon Linux 2023 |
|
Rocky Linux 9.1 or later |
|
Kylin Euler V10 SP1/SP2 |
|
UnionTech OS (UOS) V20 |
|
openEuler 22.03 LTS SP1/SP3 |
|
macOS 12 (Monterey) or later |
|
Oracle Enterprise Linux 8 or a later | x86_64 |
Ubuntu LTS 20.04 or later | x86_64 |
CentOS 8 Stream |
|
Debian 10 (Buster) or later | x86_64 |
Fedora 38 or later | x86_64 |
openSUSE Leap later than v15.5 (not including Tumbleweed) | x86_64 |
SUSE Linux Enterprise Server 15 | x86_64 |
Note:
- For Oracle Enterprise Linux, TiDB supports the Red Hat Compatible Kernel (RHCK) and does not support the Unbreakable Enterprise Kernel provided by Oracle Enterprise Linux.
- According to CentOS Linux EOL, the upstream support for CentOS Linux 7 ends on June 30, 2024. TiDB ends the support for CentOS 7 starting from the 8.4 DMR version. It is recommended to use Rocky Linux 9.1 or a later version. While the upstream support for CentOS Linux 8 ends on December 31, 2021, CentOS Stream 8 continues to be supported by the CentOS organization.
- Support for Ubuntu 16.04 will be removed in future versions of TiDB. Upgrading to Ubuntu 18.04 or later is strongly recommended.
- If you are using the 32-bit version of an operating system listed in the preceding table, TiDB is not guaranteed to be compilable, buildable or deployable on the 32-bit operating system and the corresponding CPU architecture, or TiDB does not actively adapt to the 32-bit operating system.
- Other operating system versions not mentioned above might work but are not officially supported.
Libraries required for compiling and running TiDB | Version |
---|---|
Golang | 1.23 or later |
Rust | nightly-2023-12-28 or later |
GCC | 7.x |
LLVM | 17.0 or later |
Library required for running TiDB: glibc (2.28-151.el8 version)
The following CPU architectures are supported:
- x86_64. Starting from TiDB v6.6.0, the x86-64-v2 instruction set is required.
- ARM 64
Software | Version |
---|---|
sshpass | 1.06 or later |
TiUP | 1.5.0 or later |
Note:
It is required that you deploy TiUP on the control machine to operate and manage TiDB clusters.
Software | Version |
---|---|
sshpass | 1.06 or later |
numa | 2.0.12 or later |
tar | any |
You can deploy and run TiDB on the 64-bit generic hardware server platform in the Intel x86-64 architecture or on the hardware server platform in the ARM architecture. The requirements and recommendations about server hardware configuration (ignoring the resources occupied by the operating system itself) for development, test, and production environments are as follows:
Component | CPU | Memory | Local Storage | Network | Number of Instances (Minimum Requirement) |
---|---|---|---|---|---|
TiDB | 8 core+ | 16 GB+ | Disk space requirements | Gigabit network card | 1 (can be deployed on the same machine with PD) |
PD | 4 core+ | 8 GB+ | SAS, 200 GB+ | Gigabit network card | 1 (can be deployed on the same machine with TiDB) |
TiKV | 8 core+ | 32 GB+ | SAS, 200 GB+ | Gigabit network card | 3 |
TiFlash | 32 core+ | 64 GB+ | SSD, 200 GB+ | Gigabit network card | 1 |
TiCDC | 8 core+ | 16 GB+ | SAS, 200 GB+ | Gigabit network card | 1 |
Note:
- In the test environment, the TiDB and PD instances can be deployed on the same server.
- For performance-related test, do not use low-performance storage and network hardware configuration, in order to guarantee the correctness of the test result.
- For the TiKV server, it is recommended to use NVMe SSDs to ensure faster reads and writes.
- If you only want to test and verify the features, follow Quick Start Guide for TiDB to deploy TiDB on a single machine.
- Starting from v6.3.0, to deploy TiFlash under the Linux AMD64 architecture, the CPU must support the AVX2 instruction set. Ensure that
grep avx2 /proc/cpuinfo
has output. To deploy TiFlash under the Linux ARM64 architecture, the CPU must support the ARMv8 instruction set architecture. Ensure thatgrep 'crc32' /proc/cpuinfo | grep 'asimd'
has output. By using the instruction set extensions, TiFlash's vectorization engine can deliver better performance.
Component | CPU | Memory | Hard Disk Type | Network | Number of Instances (Minimum Requirement) |
---|---|---|---|---|---|
TiDB | 16 core+ | 48 GB+ | SSD | 10 Gigabit network card (2 preferred) | 2 |
PD | 8 core+ | 16 GB+ | SSD | 10 Gigabit network card (2 preferred) | 3 |
TiKV | 16 core+ | 64 GB+ | SSD | 10 Gigabit network card (2 preferred) | 3 |
TiFlash | 48 core+ | 128 GB+ | 1 or more SSDs | 10 Gigabit network card (2 preferred) | 2 |
TiCDC | 16 core+ | 64 GB+ | SSD | 10 Gigabit network card (2 preferred) | 2 |
Monitor | 8 core+ | 16 GB+ | SAS | Gigabit network card | 1 |
Note:
- In the production environment, the TiDB and PD instances can be deployed on the same server. If you have a higher requirement for performance and reliability, try to deploy them separately.
- It is strongly recommended to configure TiDB, TiKV, and TiFlash with at least 8 CPU cores each in the production environment. To get better performance, a higher configuration is recommended.
- It is recommended to keep the size of TiKV hard disk within 4 TB if you are using PCIe SSDs or within 1.5 TB if you are using regular SSDs.
- If you deploy TiKV on a cloud provider, such as AWS, Google Cloud, or Azure, it is recommended to use cloud disks for TiKV nodes. Data on local disks might be lost if the TiKV instance crashes in the cloud environment.
Before you deploy TiFlash, note the following items:
- TiFlash can be deployed on multiple disks.
- It is recommended to use a high-performance SSD as the first disk of the TiFlash data directory to buffer the real-time replication of TiKV data. The performance of this disk should not be lower than that of TiKV, such as PCIe SSD. The disk capacity should be no less than 10% of the total capacity; otherwise, it might become the bottleneck of this node. You can deploy ordinary SSDs for other disks, but note that a better PCIe SSD brings better performance.
- It is recommended to deploy TiFlash on different nodes from TiKV. If you must deploy TiFlash and TiKV on the same node, increase the number of CPU cores and memory, and try to deploy TiFlash and TiKV on different disks to avoid interfering each other.
- The total capacity of the TiFlash disks is calculated in this way:
the data volume of the entire TiKV cluster to be replicated / the number of TiKV replicas * the number of TiFlash replicas
. For example, if the overall planned capacity of TiKV is 1 TB, the number of TiKV replicas is 3, and the number of TiFlash replicas is 2, then the recommended total capacity of TiFlash is1024 GB / 3 * 2
. You can replicate only the data of some tables. In such case, determine the TiFlash capacity according to the data volume of the tables to be replicated.
Before you deploy TiCDC, note that it is recommended to deploy TiCDC on PCIe SSD disks larger than 500 GB.
As an open-source distributed SQL database, TiDB requires the following network port configuration to run. Based on the TiDB deployment in actual environments, the administrator can open relevant ports in the network side and host side.
Component | Default Port | Description |
---|---|---|
TiDB | 4000 | the communication port for the application and DBA tools |
TiDB | 10080 | the communication port to report TiDB status |
TiKV | 20160 | the TiKV communication port |
TiKV | 20180 | the communication port to report TiKV status |
PD | 2379 | the communication port between TiDB and PD |
PD | 2380 | the inter-node communication port within the PD cluster |
TiFlash | 9000 | the TiFlash TCP service port |
TiFlash | 3930 | the TiFlash RAFT and Coprocessor service port |
TiFlash | 20170 | the TiFlash Proxy service port |
TiFlash | 20292 | the port for Prometheus to pull TiFlash Proxy metrics |
TiFlash | 8234 | the port for Prometheus to pull TiFlash metrics |
TiCDC | 8300 | the TiCDC communication port |
Monitoring | 9090 | the communication port for the Prometheus service |
Monitoring | 12020 | the communication port for the NgMonitoring service |
Node_exporter | 9100 | the communication port to report the system information of every TiDB cluster node |
Blackbox_exporter | 9115 | the Blackbox_exporter communication port, used to monitor the ports in the TiDB cluster |
Grafana | 3000 | the port for the external Web monitoring service and client (Browser) access |
Alertmanager | 9093 | the port for the alert web service |
Alertmanager | 9094 | the alert communication port |
Component | Disk space requirement | Healthy disk usage |
---|---|---|
TiDB |
|
Lower than 90% |
PD | At least 20 GB for the data disk and for the log disk, respectively | Lower than 90% |
TiKV | At least 100 GB for the data disk and for the log disk, respectively | Lower than 80% |
TiFlash | At least 100 GB for the data disk and at least 30 GB for the log disk, respectively | Lower than 80% |
TiUP |
|
N/A |
Ngmonitoring |
|
N/A |
TiDB relies on Grafana to provide visualization of database metrics. A recent version of Microsoft Edge, Safari, Chrome or Firefox with Javascript enabled is sufficient.
The preceding TiFlash software and hardware requirements are for the coupled storage and compute architecture. Starting from v7.0.0, TiFlash supports the disaggregated storage and compute architecture. In this architecture, TiFlash is divided into two types of nodes: the Write Node and the Compute Node. The requirements for these nodes are as follows:
- Software: remain the same as the coupled storage and compute architecture, see OS and platform requirements.
- Network port: remain the same as the coupled storage and compute architecture, see Network.
- Disk space:
- TiFlash Write Node: it is recommended to configure at least 200 GB of disk space, which is used as a local buffer when adding TiFlash replicas and migrating Region replicas before uploading data to Amazon S3. In addition, an object storage compatible with Amazon S3 is required.
- TiFlash Compute Node: it is recommended to configure at least 100 GB of disk space, which is mainly used to cache the data read from the Write Node to improve performance. The cache of the Compute Node might be fully used, which is normal.
- CPU and memory requirements are described in the following sections.
Component | CPU | Memory | Local Storage | Network | Number of Instances (Minimum Requirement) |
---|---|---|---|---|---|
TiFlash Write Node | 16 cores+ | 32 GB+ | SSD, 200 GB+ | Gigabit Ethernet | 1 |
TiFlash Compute Node | 16 cores+ | 32 GB+ | SSD, 100 GB+ | Gigabit Ethernet | 0 (see the following note) |
Component | CPU | Memory | Disk Type | Network | Number of Instances (Minimum Requirement) |
---|---|---|---|---|---|
TiFlash Write Node | 32 cores+ | 64 GB+ | 1 or more SSDs | 10 Gigabit Ethernet (2 recommended) | 1 |
TiFlash Compute Node | 32 cores+ | 64 GB+ | 1 or more SSDs | 10 Gigabit Ethernet (2 recommended) | 0 (see the following note) |
Note:
You can use deployment tools such as TiUP to quickly scale in or out the TiFlash Compute Node, within the range of
[0, +inf]
.