Your company here — click to reach over 10,000 unique daily visitors

tmt - Man Page

Test Management Tool


The tmt tool provides a user-friendly way to work with tests. You can comfortably create new tests, safely and easily run tests across different environments, review test results, debug test code and enable tests in the CI using a consistent and concise config.

The python module and command-line tool implement the Metadata Specification which allows storing all needed test execution data directly within a git repository. Together with possibility to reference remote repositories it makes it easy to share test coverage across projects and distros.

The Flexible Metadata Format fmf is used to store data in both human and machine readable way close to the source code. Thanks to inheritance and elasticity metadata are organized in the structure efficiently, preventing unnecessary duplication.


There are several metadata levels defined by the specification:

Core attributes such as summary or description which are common across all levels are defined by the special L0 metadata.

Tests, or L1 metadata, define attributes which are closely related to individual test cases such as test script, framework, directory path where the test should be executed, maximum test duration or packages required to run the test.

Plans, also called L2 metadata, are used to group relevant tests and enable them in the CI. They describe how to discover tests for execution, how to provision the environment, how to prepare it for testing, how to execute tests and report test results.

Stories, which implement the L3 metadata, can be used to track implementation, test and documentation coverage for individual features or requirements. Thanks to this you can track everything in one place, including the project implementation progress.


Command line usage is straightforward:

tmt command [options]


Let's see which tests, plans and stories are available:


Initialize the metadata tree in the current directory, optionally with example content based on templates:

tmt init
tmt init --template base

Run all or selected steps for each plan:

tmt run
tmt run discover
tmt run prepare execute

See any failures from a previous run:

tmt run --last report -vvv

Show test output while running:

tmt run -vvv

List tests, show details, check against the specification:

tmt tests ls
tmt tests show
tmt tests lint

Create a new test, import test metadata from other formats:

tmt test create
tmt test import

List plans, show details, check against the specification:

tmt plans ls
tmt plans show
tmt plans lint

List stories, check details, show coverage status:

tmt stories ls
tmt stories show
tmt stories coverage

Many commands support regular expression filtering and other specific options:

tmt stories ls cli
tmt stories show create
tmt stories coverage --implemented

Check help message of individual commands for the full list of available options.


Here is the list of the most frequently used commands and options.


The run command is used to execute test steps. By default all test steps are run. See the L2 Metadata specification for detailed description of individual steps. Here is a brief overview:


Gather information about test cases to be executed.


Provision an environment for testing or use localhost.


Prepare the environment for testing.


Run tests using the specified executor.


Provide test results overview and send reports.


Perform the finishing tasks and clean up provisioned guests.


Manage tests (L1 metadata). Check available tests, inspect their metadata, gather old metadata from various sources and stored them in the new fmf format.


List available tests.


Show test details.


Check tests against the L1 metadata specification.


Create a new test based on given template.


Convert old test metadata into the new fmf format.


Manage test plans (L2 metadata). Search for available plans. Explore detailed test step configuration.


List available plans.


Show plan details.


Check plans against the L2 metadata specification.


Manage user stories. Check available user stories. Explore coverage (test, implementation, documentation).


List available stories.


Show story details.


Show code, test and docs coverage for given stories.


Export selected stories into desired format.


Various utility options.

--root PATH

Path to the metadata tree, current directory used by default.


Print additional information.


Turn on debugging output.

Check help message of individual commands for the full list of available options.


The main tmt package provides the core features with a minimal set of dependencies:

sudo dnf install tmt

In order to enable additional functionality, such as particular provision or report plugins, install the respective subpackage:

sudo dnf install tmt+test-convert
sudo dnf install tmt+provision-container
sudo dnf install tmt+provision-virtual

If you don't care about disk space and want to have all available features right at hand install everything:

sudo dnf install tmt+all

For CentOS and RHEL, first make sure that you have available the EPEL repository. You might also have to enable additional repositories:

sudo dnf config-manager --enable crb         # CentOS 9
sudo dnf config-manager --enable rhel-CRB    # RHEL 9
sudo dnf install https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm

sudo dnf install tmt

For plugins which cannot work outside of VPN and so live within its walls you need to enable the internal copr repository first. Then you can install either everything or only those you need:

sudo dnf install tmt-redhat-all
sudo dnf install tmt-redhat-*

Impatient to try the fresh features as soon as possible? Install the latest greatest version from the copr repository:

sudo dnf copr enable @teemtee/tmt
sudo dnf install tmt

Not sure, just want to try out how it works? Experiment safely and easily inside a container:

podman run -it --rm quay.io/testing-farm/tmt bash
podman run -it --rm quay.io/testing-farm/tmt-all bash

When installing using pip you might need to install additional packages on your system:

sudo dnf install gcc redhat-rpm-config
sudo dnf install {python3,libvirt,krb5,libpq}-devel
pip install --user tmt

On other distributions than Fedora or RHEL the package names might be different. For example on Ubuntu to install all packages to have provision plugins working:

sudo apt install libkrb5-dev pkg-config libvirt-dev genisoimage qemu-kvm libvirt-daemon-system
pip install --user "tmt[provision]"

Note: You can omit the --user flag if in a virtual environment.

Shell Completion

The rpm package includes a system wide script which enables the command line completion for bash so no additional config should be needed. If you use a different installation method or prefer another shell, see the instructions below.

For Bash, add this to ~/.bashrc:

eval "$(_TMT_COMPLETE=source_bash tmt)"

For Zsh, add this to ~/.zshrc:

eval "$(_TMT_COMPLETE=source_zsh tmt)"

For Fish, add this to ~/.config/fish/completions/tmt.fish:

eval (env _TMT_COMPLETE=source_fish tmt)

Open a new shell to enable completion. Or run the eval command directly in your current shell to enable it temporarily.

This is however run every time you start a shell which can cause some delay. To speed it up, write the generated script to a file and then source it from your shell's configuration file. All of this can be achieved using tmt setup completion command. By default, it outputs the completion script to the terminal but it can also add it to your ~/.bashrc or ~/.zshrc using the --install option:

tmt setup completion {bash, zsh, fish} --install

Exit Codes

The following exit codes are returned from tmt run. Note that you can use the --quiet option to completely disable output and only check for the exit code.


At least one test passed, there was no fail, warn or error.


There was a fail or warn identified, but no error.


Errors occurred during test execution.


No test results found.


Tests were executed, and all reported the skip result.


The list of available environment variables which can be used to adjust the execution.

Command Variables

The following environment variables can be used to modify behaviour of the tmt command.


Enable the desired debug level. Most of the commands support levels from 1 to 3. However, some of the plugins go even deeper when needed.


Path to a directory with additional plugins. Multiple paths separated with the : character can be provided as well.


Path to an alternative directory with config files. By default ~/.config/tmt is used.


Path to root directory containing run workdirs. Defaults to /var/tmp/tmt.


Disable colors in the output, both the actual output and logging messages. Output only plain, non-colored text.

Two variables are accepted, one with the usual TMT_ prefix, but tmt accepts also NO_COLOR to support the NO_COLOR effort, see https://no-color.org/ for more information.


Enforce colors in the output, both the actual output and logging messages. Might come handy when tmt's output streams are not terminal-like, yet its output would be displayed by tools with ANSI color support. This is often the case of various CI systems.

Note that TMT_FORCE_COLOR takes priority over NO_COLOR and TMT_NO_COLOR. If user tries both to disable and enable colorization, output would be colorized.


By default, when tmt reports an error, the corresponding traceback is not printed out. When TMT_SHOW_TRACEBACK is set to any string except 0, traceback would be printed out. When set to full, traceback would list also local variables in each stack frame.


By default, the output width of commands like tmt * show is constrained to 79 characters. Set this variable to an integer to change the limit.


Variable pairs used to provide credentials to clone git repositories. Suffix identifies the pair and determines the order in which URL regexp is tried.

The TMT_GIT_CREDENTIALS_URL_<suffix> contains regexp to search against url to clone. For first successful search the content of the TMT_GIT_CREDENTIALS_VALUE_<suffix> variable is used as the credential value. When it is set to an empty string, unmodified url is used.

Example usage:

GitLab credentials need to contain nonempty username followed by colon and token value:


GitHub credentials contain just the token value:


The maximum number of retries to clone a git repository if it fails. By default, 3 attempts are done.


The interval (in seconds) to retry cloning a git repository again, 10 seconds by default.


Overall maximum time in seconds to clone a git repository. By default, the limit is not set.


How many seconds to wait for a guest to boot. Applies to provision plugins that control the guest creation, e.g. virtual. By default, it is 2 minutes.


How many seconds to wait for a connection to succeed after guest boot. By default, it is 2 minutes.


How many seconds to wait for a connection to succeed after guest reboot. By default, it is 10 minutes.


Every environment variable in this format would be treated as an SSH option, and passed to the -o option of ssh command. See man 5 ssh_config for the list of all options.

The environment variable name would be converted into an SSH option, e.g. export TMT_SSH_CONNECTION_ATTEMPTS=5 would become -oConnectionAttempts=5. export TMT_SSH_ConnectionAttempts=5 would also be accepted.

SSH options provided via environment variables will be overruled by guest-specific ssh-options key.


Link to test artifacts provided for report plugins.

Step Variables

The following environment variables are provided to the environment during prepare, execute and finish steps:


The full path of the working directory where the metadata tree is copied. This usually contains the whole git repository from which tests have been executed.


Path to the common directory used for storing logs and other artifacts related to the whole plan execution. It is pulled back from the guest and available for inspection after the plan is completed.


Path to the file containing environment variables that should be sourced after prepare and execute steps. These variables will be accessible for all subsequent steps and have lower priority than variables specified by the environment key, environment-file key, or the command line. Variables inside the file have to be in the format of NAME=VALUE and each variable should be on a separate line.

Example of the file content:


The version of tmt.

Test Variables

The following environment variables are provided to the test during the execution:


The test name, as a resolved FMF object name starting with / from the root of the hierarchy.


Path to the directory where test can store logs and other artifacts generated during its execution. These will be pulled back from the guest and available for inspection after the test execution is finished.


The serial number of running test in the whole plan. Each test is assigned its own serial number.


The iteration ID is a combination of a unique run ID and the test serial number. The value is different for each new test execution.


Path to a YAML-formatted file with test metadata collected during the discover step.


Path to directory with downloaded and extracted sources if the dist-git-source option was used in the discover step.


During the test execution the tmt-reboot command can be used to request reboot of the guest. This variable contains number of reboots which already happened during the test. Value is set to 0 if no reboot occurred.

In order to keep backward-compatibility with older tests, rhts-reboot and rstrnt-reboot commands are supported for requesting the reboot, variables REBOOTCOUNT and RSTRNT_REBOOTCOUNT contain number of reboots as well.


This variable contains number of times the test was restarted. Such restarts may be consequence of guest reboot, in which case TMT_REBOOT_COUNT gets incremented as well, or test crashed and has been restarted. Value is set to 0 when the test starts for the first time.


Paths of files describing existing guests, their roles and the guest on which the test is running. Format of these files is described in the Guest Topology Format section of the plan specification.


Path to a file storing the test process pid and path to its reboot-request file, separated by a space. The "LOCK" variable then holds path to a locking file which must be acquired before making any changes to the pid file.


By default, the test pidfile file is stored in /var/tmp directory. If specified, the directory in this variable would be used instead. The directory permissions should follow the pattern of temporary directory permissions, e.g. chmod 1777, to allow access to users with all privilege levels.

Plugin Variables

Each plugin option can be also specified via environment variable. Variables follow a naming scheme utilizing plugin name, step it belongs to, and the option name:


All values are upper-cased, with dashes (-) replaced by underscores (_).

For example, an execute plugin "tmt" would run with verbosity equal to -vvv:

TMT_PLUGIN_EXECUTE_TMT_VERBOSE=3 tmt run ... execute -h tmt ...

Command-line takes precedence over environment variables, therefore -v would undo the effect of environment variable, and reduce verbosity to one level only:

TMT_PLUGIN_EXECUTE_TMT_VERBOSE=3 tmt run ... execute -h tmt -v ...

Environment variables - just like command-line options - take precedence over values stored in files. For example, consider the following discover step:

    how: fmf
    url: https://example.org/

The following commands would override the URL:

tmt run ... discover -h fmf --url https://actual.org/ ...

TMT_PLUGIN_DISCOVER_FMF_URL=https://actual.org/ tmt run ...

For setting flag-like option, 0 and 1 are the expected value. For example, an interactive mode would be enabled in this run:

TMT_PLUGIN_EXECUTE_TMT_INTERACTIVE=1 tmt run ... execute -h tmt ...

The following applies to situations when a plugin is specified on the command line only. Keys of plugins specified in fmf files would not be modified. This is a limit of the current implementation, and will be addressed in the future:

# Here the verbosity will not be increased since the plugin is
# not mentioned on the command line:

# Here the environment variable will take effect:
$ TMT_PLUGIN_DISCOVER_FMF_VERBOSE=2 tmt run -a discover -h fmf ...

Regular Expressions

Many specification keys and command line options accept regular expressions, e.g. to filter a set of tests, tmt run ... test --name '^/foo'. Whenever tmt works with regular expressions, the following rules apply.

Python implementation

Since tmt is implemented in Python, Python's re package is used for handling regular expressions. It comes with Python standard library, and it is widely used and well documented. For the deep dive into supported syntax, see Regular Expression Syntax section, for a gentler introduction there is a Regular Expression HOWTO.

Search versus match

There are two ways how to check whether a regular expression matches a string: "match" or "search":

  • in the "match" mode, the pattern must match from the very beginning of the string. For example, foo.ar would match foobar but not /foobar. It works in a "starts with" fashion.
  • on the other hand, the "search" mode is more similar to "contains" approach, and allows pattern to match anywhere in the string. In this mode, foo.ar would match both foobar and /foobar.

In both modes, any characters may follow the matching pattern, e.g. regular expression foo.bar is the same as foo.bar.*.

tmt sticks to the "search" mode by default. But, to improve user experience, some keys and command line options do use the "match" mode. Such keys and options will note this in their documentation or help texts.


Petr Šplíchal, Miro Hrončok, Alexander Sosedkin, Lukáš Zachar, Petr Menšík, Leoš Pol, Miroslav Vadkerti, Pavel Valena, Jakub Heger, Honza Horák, Rachel Sibley, František Nečas, Michal Ruprich, Martin Kyral, Miloš Prchlík, Tomáš Navrátil, František Lachman, Patrik Kis, Ondrej Mosnáček, Andrea Ficková, Denis Karpelevich, Michal Srb, Jan Ščotka, Artem Zhukov, Vinzenz Feenstra, Inessa Vasilevskaya, Štěpán Němec, Robin Hack, Yulia Kopkova, Ondrej Moriš, Martin Zelený, Karel Šrot, František Zatloukal, Simon Walter, Petr Matyáš, Yariv Rachmani, Pavel Cahyna, Martin Litwora, Brian Grech, Vojtěch Eichler, Philip Daly, Vector Li, Evgeny Fedin, Guy Inger, Adrián Tomašov, Jan Havlín, Lukáš Kotek, Daniel Diblík, Laura Barcziova, Marián Konček, Marcin Sobczyk, Jiří Jabůrek, Huijing Hei, Tibor Dudlák, Jan Macků, Filip Vágner, Martin Hoyer, Iveta Česalová, Yi Zhang, Zhaojuan Guo, Natália Bubáková, Michal Josef Špaček, Jiří Popelka, Matej Focko, Yulia Kopkova, Tomáš Bajer, Carlos Rodriguez-Fernandez, James Molet, Cristian Le, Lili Nie, Martin Čermák, Michael Vogt and Qinghua Cheng.


September 2019