Skip to content

Latest commit

 

History

History
125 lines (74 loc) · 5.96 KB

troubleshoot-tidb-cluster.md

File metadata and controls

125 lines (74 loc) · 5.96 KB
Raw
title summary aliases
TiDB Cluster Troubleshooting Guide
Learn how to diagnose and resolve issues when you use TiDB.
/docs/dev/troubleshoot-tidb-cluster/
/docs/dev/how-to/troubleshoot/cluster-setup/

TiDB Cluster Troubleshooting Guide

You can use this guide to help you diagnose and solve basic problems while using TiDB. If your problem is not resolved, please collect the following information and create an issue:

  • The exact error message and the operations while the error occurs
  • The state of all the components
  • The error/fatal/panic information in the log of the component that reports the error
  • The configuration and deployment topology
  • The TiDB component related issue in dmesg

For other information, see Frequently Asked Questions (FAQ).

Cannot connect to the database

  1. Make sure all the services are started, including tidb-server, pd-server, and tikv-server.

  2. Use the ps command to check if all the processes are running.

    • If a certain process is not running, see the following corresponding sections to diagnose and solve the issue.
    • If all the processes are running, check the tidb-server log to see if the following messages are displayed:
      • InformationSchema is out of date: This message is displayed if the tikv-server cannot be connected. Check the state and log of pd-server and tikv-server.
      • panic: This message is displayed if there is an issue with the program. Please provide the detailed panic log and create an issue.

  3. If the data is cleared and the services are re-deployed, make sure that:

    • All the data in tikv-server and pd-server are cleared. The specific data is stored in tikv-server and the metadata is stored in pd-server. If only one of the two servers is cleared, the data will be inconsistent.
    • After the data in pd-server and tikv-server are cleared and the pd-server and tikv-server are restarted, the tidb-server must be restarted too. The cluster ID is randomly allocated when the pd-server is initialized. So when the cluster is re-deployed, the cluster ID changes and you need to restart the tidb-server to get the new cluster ID.

Cannot start tidb-server

See the following for the situations when the tidb-server cannot be started:

  • Error in the startup parameters.

    See the TiDB configuration and options.

  • The port is occupied.

    Use the lsof -i:port command to show all the networking related to a given port and make sure the port to start the tidb-server is not occupied.

  • Cannot connect to pd-server.

    • Check if the network between TiDB and PD is running smoothly, including whether the network can be pinged or if there is any issue with the Firewall configuration.
    • If there is no issue with the network, check the state and log of the pd-server process.

Cannot start tikv-server

See the following for the situations when the tikv-server cannot be started:

  • Error in the startup parameters: See the TiKV configuration and options.

  • The port is occupied: Use the lsof -i:port command to show all the networking related to a given port and make sure the port to start the tikv-server is not occupied.

  • Cannot connect to pd-server.

    • Check if the network between TiDB and PD is running smoothly, including whether the network can be pinged or if there is any issue with the Firewall configuration.

    • If there is no issue with the network, check the state and log of the pd-server process.

  • The file is occupied.

    Do not open two TiKV files on one database file directory.

Cannot start pd-server

See the following for the situations when the pd-server cannot be started:

  • Error in the startup parameters.

    See the PD configuration and options.

  • The port is occupied.

    Use the lsof -i:port command to show all the networking related to a given port and make sure the port to start the pd-server is not occupied.

The TiDB/TiKV/PD process aborts unexpectedly

  • Is the process started on the foreground? The process might exit because the client aborts.

  • Is nohup+& run in the command line? This might cause the process to abort because it receives the hup signal. It is recommended to write and run the startup command in a script.

TiDB panic

Please provide the panic log and create an issue.

The connection is rejected

Make sure the network parameters of the operating system are correct, including but not limited to:

  • The port in the connection string is consistent with the tidb-server starting port.
  • The firewall is configured correctly.

Open too many files

Before starting the process, make sure the result of ulimit -n is large enough. It is recommended to set the value to unlimited or larger than 1000000.

Database access times out and the system load is too high

First, check the slow query log and see if it is because of some inappropriate SQL statement.

If you failed to solve the problem, provide the following information:

  • The deployment topology

    • How many tidb-server/pd-server/tikv-server instances are deployed?
    • How are these instances distributed in the machines?

  • The hardware configuration of the machines where these instances are deployed:

    • The number of CPU cores
    • The size of the memory
    • The type of the disk (SSD or Hard Drive Disk)
    • Are they physical machines or virtual machines?
  • Are there other services besides the TiDB cluster?
  • Are the pd-servers and tikv-servers deployed separately?
  • What is the current operation?
  • Check the CPU thread name using the top -H command.
  • Are there any exceptions in the network or IO monitoring data recently?