Awesome

dbbench

Table of Contents

• Description
• Example
• Installation
• Supported Databases
• Usage
• Custom Scripts
• Troubeshooting
• Development
• Acknowledgements

Description

dbbench is a simple tool to benchmark or stress test databases. You can use the simple built-in benchmarks or run your own queries.

Attention: This tool comes with no warranty. Don't run it on production databases.

Example

$ dbbench postgres --user postgres --pass example --iter 100000
inserts 6.199670776s    61996   ns/op
updates 7.74049898s     77404   ns/op
selects 2.911541197s    29115   ns/op
deletes 5.999572479s    59995   ns/op
total: 22.85141994s

Installation

Precompiled Binaries

Binaries are available for all major platforms. See the releases page. Unfortunately, cgo is disabled for these builds, which means there is no SQLite support (#1).

Homebrew

Using the Homebrew package manager for macOS:

brew install sj14/tap/dbbench

Manually

It's also possible to install the current development snapshot with go get (not recommended):

go get -u github.com/sj14/dbbench/cmd/dbbench

Docker

docker run ghcr.io/sj14/dbbench:latest

Check here for available images.

Supported Databases / Driver

Usage

Available subcommands:
        cassandra|cockroach|mssql|mysql|postgres|sqlite
        Use 'subcommand --help' for all flags of the specified command.
Generic flags for all subcommands:
      --clean            only cleanup benchmark data, e.g. after a crash
      --iter int         how many iterations should be run (default 1000)
      --noclean          keep benchmark data
      --noinit           do not initialize database and tables, e.g. when only running own script
      --run string       only run the specified benchmarks, e.g. "inserts deletes" (default "all")
      --script string    custom sql file to execute
      --sleep duration   how long to pause after each single benchmark (valid units: ns, us, ms, s, m, h)
      --threads int      max. number of green threads (iter >= threads > 0) (default 25)
      --version          print version information

Custom Scripts

You can run your own SQL statements with the --script flag. You can use the auto-generate tables. Beware the file size as it will be completely loaded into memory.

The script must contain valid SQL statements for your database.

There are some built-in variables and functions which can be used in the script. It's using the golang template engine which uses the delimiters {{ and }}. Functions are executed with the call command and arguments are passed after the function name.

Benchmark Settings

A new benchmark is created with the \benchmark keyword, followed by either once or loop. Optional parameters can be added afterwards in the same line.

The the usage description and the example subsection for more information.

Statement Substitutions

Example

Exemplary sqlite_bench.sql file:

-- Create table
\benchmark once \name init
CREATE TABLE dbbench_simple (id INT PRIMARY KEY, balance DECIMAL);

-- How long takes an insert and delete?
\benchmark loop \name single
INSERT INTO dbbench_simple (id, balance) VALUES({{.Iter}}, {{call .RandInt64}});
DELETE FROM dbbench_simple WHERE id = {{.Iter}}; 

-- How long takes it in a single transaction?
\benchmark loop \name batch
BEGIN TRANSACTION;
INSERT INTO dbbench_simple (id, balance) VALUES({{.Iter}}, {{call .RandInt64}});
DELETE FROM dbbench_simple WHERE id = {{.Iter}}; 
COMMIT;

-- Delete table
\benchmark once \name clean
DROP TABLE dbbench_simple;

In this script, we create and delete the table manually, thus we will pass the --noinit and --noclean flag, which would otherwise create this default table for us:

dbbench sqlite --script scripts/sqlite_bench.sql --iter 5000 --noinit --noclean

output:

(once) init:    3.404784ms      3404784 ns/op
(loop) single:  10.568390874s   2113678 ns/op
(loop) batch:   5.739021596s    1147804 ns/op
(once) clean:   1.065703ms      1065703 ns/op
total: 16.312319959s

Troubleshooting

Error message

failed to insert: UNIQUE constraint failed: dbbench_simple.id

Description The previous data wasn't removed (e.g. because the benchmark was canceled). Try to run the same command again, but with the --clean flag attached, which will remove the old data. Then run the original command again.

Error message

failed to create table: Binary was compiled with 'CGO_ENABLED=0', go-sqlite3 requires cgo to work. This is a stub

Description
Currently, the released binary builds don't contain SQLite support. You have to compile dbbench manually, either from the particular release source code (recommended) or from the current master branch (not recommended).

Development

Below are some examples how to run different databases and the equivalent call of dbbench for testing/developing.

Cassandra

docker run --name dbbench-cassandra -p 9042:9042 -d cassandra:latest

dbbench cassandra

CockroachDB

# port 8080 is the webinterface (optional)
docker run --name dbbench-cockroach -d -p 26257:26257 -p 8080:8080 cockroachdb/cockroach:latest start --insecure

dbbench cockroach

Microsoft SQL Server

docker run -e 'ACCEPT_EULA=Y' -e 'SA_PASSWORD=yourStrong(!)Password' -p 1433:1433 -d microsoft/mssql-server-linux

dbbench mssql -user sa -pass 'yourStrong(!)Password'

MariaDB

docker run --name dbbench-mariadb -p 3306:3306 -d -e MYSQL_ROOT_PASSWORD=root mariadb

dbbench mariadb

MySQL

docker run --name dbbench-mysql -p 3306:3306 -d -e MYSQL_ROOT_PASSWORD=root mysql

dbbench mysql

PostgreSQL

docker run --name dbbench-postgres -p 5432:5432 -d postgres

dbbench postgres --user postgres --pass example

ScyllaDB

docker run --name dbbench-scylla -p 9042:9042 -d scylladb/scylla

dbbench scylla

SQLite

dbbench sqlite

TiDB

git clone https://github.com/pingcap/tidb-docker-compose.git

cd tidb-docker-compose && docker-compose pull

docker-compose up -d

dbbench tidb --pass '' --port 4000

Acknowledgements

Thanks to the authors of Go and those of the directly and indirectly used libraries, especially the driver developers. It wouldn't be possible without all your work.

This tool was highly inspired by the snippet from user Fale and the tool pgbench. Later, also inspired by MemSQL's dbbench which had the name and a similar idea before.