Skip to content

vitessio/vt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

VT Utilities

The vt binary encapsulates several utility tools for Vitess, providing a comprehensive suite for testing, summarizing, and query analysis.

Tools Included

  • vt test: A testing utility using the same test files as the MySQL Test Framework. It compares the results of identical queries executed on both MySQL and Vitess (vtgate), helping to ensure compatibility.
  • vt keys: A utility that analyzes query logs and provides information about queries, tables, joins, and column usage.
  • vt transactions: A tool that analyzes query logs to identify transaction patterns and outputs a JSON report detailing these patterns.
  • vt trace: A tool that generates execution traces for queries without comparing against MySQL. It helps analyze query behavior and performance in Vitess environments.
  • vt summarize: A tool used to summarize or compare trace logs or key logs for easier human consumption.
  • vt dbinfo: A tool that provides information about the database schema, including row counts, useful column attributes and relevant subset of global variables.

Installation

You can install vt using the following command:

go install github.com/vitessio/vt/go/vt@latest

Testing Methodology

To verify compatibility and correctness, the testing strategy involves running identical queries on both MySQL and vtgate, followed by a comparison of results. The process includes:

  1. Query Execution: Each test query is executed on both MySQL and vtgate.
  2. Result Comparison: The returned data, result set structure (column types, order), and errors are compared.
  3. Error Handling: Any errors are checked to ensure vtgate produces the same error types as MySQL.

This dual-testing strategy ensures high confidence in vtgate's compatibility with MySQL.

Sharded Testing Strategy

Vitess operates in a sharded environment, presenting unique challenges, especially during schema changes (DDL). The vt test tool handles these by converting DDL statements into VSchema commands.

Here's an example of running vt test:

vt test --sharded t/basic.test  # Runs tests on a sharded database

Custom schemas and configurations can be applied using directives. Run vt test --help, and check out directives.test for more examples.

Tracing and Query Analysis

Vitess provides two main approaches for tracing query execution:

Comparative Tracing with vt test

vt test can generate traces while comparing behavior with MySQL using the --trace-file flag:

vt test --sharded --trace-file=trace-log.json t/tpch.test

Standalone Tracing with vt trace

vt trace focuses solely on analyzing query execution in Vitess without MySQL comparison:

# With VSchema and backup initialization
vt trace --vschema=t/vschema.json --backup-path=/path/to/backup --number-of-shards=4 t/tpch.test > trace-log.json

vt trace accepts most of the same configuration flags as vt test, including:

  • --sharded: Enable auto-sharded mode - uses primary keys as sharding keys. Not a good idea for a production environment, but can be used to ensure that all queries work in a sharded environment.
  • --vschema: Specify the VSchema configuration
  • --backup-path: Initialize from a backup
  • --number-of-shards: Specify the number of shards to bring up
  • Other database configuration flags

Both vt trace and vt keys support different input file formats through the --input-type flag:

Example using different input types:

# Analyze SQL file or slow query log
vt trace slow-query.log > trace-log.json

# Analyze MySQL general query log
vt trace --input-type=mysql-log general-query.log > trace-log.json

# Analyze VTGate query log
vt trace --input-type=vtgate-log vtgate-querylog.log > trace-log.json

Both types of trace logs can be analyzed using vt summarize:

vt summarize trace-log.json  # Summarize a single trace
vt summarize trace-log1.json trace-log2.json  # Compare two traces

Key Analysis Workflow

vt keys analyzes query logs and outputs detailed information about tables, columns usage and joins in queries. This data can be summarized using vt summarize. Here's a typical workflow:

  1. Run vt keys to analyze queries:

    # Analyze an SQL file or slow query log
    vt keys slow-query.log > keys-log.json
    
    # Analyze a MySQL general query log
    vt keys --input-type=mysql-log general-query.log > keys-log.json
    
    # Analyze VTGate query log
    vt trace --input-type=vtgate-log vtgate-querylog.log > trace-log.json

This command generates a keys-log.json file that contains a detailed analysis of table and column usage from the queries.

  1. Summarize the keys-log using vt summarize:

    vt summarize keys-log.json

    This command summarizes the key analysis, providing insight into which tables and columns are used across queries, and how frequently they are involved in filters, groupings, and joins.
    Here is an example summary report.

    If you have access to the running database, you can use vt dbinfo > dbinfo.json and pass it to summarize so that the analysis can take into the account the additional information from the database schema and configuration:

    vt summarize keys-log.json dbinfo.json

Transaction Analysis with vt transactions

The vt transactions command is designed to analyze query logs and identify patterns of transactional queries. It processes the logs to find sequences of queries that form transactions and outputs a JSON report summarizing these patterns. Read more about how to use and how to read the output in the vt transactions documentation.

Using --backup-path Flag

The --backup-path flag allows vt test and vt trace to initialize tests from a database backup rather than an empty database. This is particularly helpful when verifying compatibility during version upgrades or testing stateful operations.

Example:

vt test --backup-path /path/to/backup -vschema t/vschema.json t/basic.test

Contributing

We welcome contributions in the following areas:

  • Writing documentation on how to use the framework
  • Triaging issues
  • Submitting new test cases
  • Fixing bugs in the test framework
  • Adding features from the MySQL test framework that are missing in this implementation

After cloning the repo, make sure to run

make install-hooks

to install the pre-commit hooks.

License

Vitess Tester is licensed under the Apache 2.0 license. See the LICENSE file for more information.

Acknowledgments

Vitess Tester started as a fork from pingcap/mysql-tester. We thank the original authors for their foundational work.