stestr user manual

Overview

stestr is an application for running and tracking test results. Any test run that can be represented as a subunit stream can be inserted into a repository. However, the test running mechanism assumes python is being used. It is originally forked from the testrepository project so the usage is similar.

A typical basic example workflow is:

# Create a store to manage test results in.
$ stestr init
# Do a test run
$ stestr run

Most commands in testr have comprehensive online help, and the commands:

$ stestr --help
$ stestr [command] --help

Will be useful to explore the system.

Configuration

stestr is configured with a config file that tells stestr some basic information about how to run tests. By default the config file needs to be .stestr.conf in the same directory that stestr is run from. However, the --config/-c CLI argument can specify an alternate path for it. stestr also includes an online help for all the options that can be set in the config file. It is in the top of the output of:

$ stestr run --help

However, the 2 most important options in the stestr config file are test_path and top_dir. These 2 options are used to set the unittest discovery options for stestr. (test_path is the same as –start-directory and top_dir is the same as –top-level-directory in the doc) Only test_path is a required field in the config file, if top_dir is not specified it defaults to ‘./’. It’s also worth noting that shell variables for these 2 config options (and only these 2 options) are expanded on platforms that have a shell. This enables you to have conditional discovery paths based on your environment.

For example, having a config file like:

[DEFAULT]
test_path=${TEST_PATH:-./foo/tests}

will let you override the discovery start path using the TEST_PATH environment variable.

There is also an option to specify all the options in the config file via the CLI. This way you can run stestr directly without having to write a config file and manually specify the test_path like above with the --test-path/-t CLI argument.

Running tests

To run tests the stestr run command is used. By default this will run all tests discovered using the discovery parameters in the stestr config file.

If you’d like to avoid the overhead of test discovery and just manually execute a single test (test class, or module) you can do this using the --no-discover/-n option. For example:

stestr run --no-discover project.tests.test_foo.TestFoo

you can also give it a file path and stestr will convert that to the proper python path under the covers. (assuming your project don’t manually mess with import paths) For example:

stestr run --no-discover project/tests/test_foo.py

will also bypass discovery and directly call subunit.run on the module specified.

Test Selection

Arguments passed to stestr run are used to filter test ids that will be run. stestr will perform unittest discovery to get a list of all test ids and then apply each argument as a regex filter. Tests that match any of the given filters will be run. For example, if you called stestr run foo bar this will only run the tests that have a regex match with foo or a regex match with bar.

stestr allows you do to do simple test exclusion via passing a rejection/black regexp:

$ stestr --back-regex 'slow_tests|bad_tests'

stestr also allow you to combine these arguments:

$ stestr --back-regexp 'slow_tests|bad_tests' ui\.interface

Here first we selected all tests which matches to ui\.interface, then we are dropping all test which matches slow_tests|bad_tests from the final list.

stestr also allows you to specify a blacklist file to define a set of regexes to exclude. You can specify a blacklist file with the --blacklist-file/-b option, for example:

$ stestr --blacklist_file $path_to_file

The format for the file is line separated regex, with ‘#’ used to signify the start of a comment on a line. For example:

# Blacklist File
^regex1 # Excludes these tests
.*regex2 # exclude those tests

The regexp used in the blacklist file or passed as argument, will be used to drop tests from the initial selection list. It will generate a list which will exclude any tests matching ‘^regex1’ or ‘.*regex2’. If a blacklist file is used in conjunction with the normal filters then the regex filters passed in as an argument regex will be used for the initial test selection, and the exclusion regexes from the blacklist file on top of that.

The dual of the blacklist file is the whitelist file which will include any tests matching the regexes in the file. You can specify the path to the file with --whitelist_file/-w, for example:

$ stestr --whitelist_file $path_to_file

The format for the file is more or less identical to the blacklist file:

# Whitelist File
^regex1 # Include these tests
.*regex2 # include those tests

However, instead of excluding the matches it will include them.

It’s also worth noting that you can use the test list option to dry run any selection arguments you are using. You just need to use –list/-l with your selection options to do this, for example:

$ stestr list 'regex3.*' --blacklist_file blacklist.txt

This will list all the tests which will be run by stestr using that combination of arguments.

Combining Test Results

There is sometimes a use case for running a single test suite split between multiple invocations of the stestr run command. For example, running a subset of tests with a different concurrency. In these cases you can use the --combine flag on stestr run`. When this flag is specified stestr will append the subunit stream from the test run into the most recent entry in the repository.

Alternatively, you can manually load the test results from a subunit stream into an existing test result in the repository using the --id/-i flag on the stestr load command. This will append the results from the input subunit stream to the specified id.

Running previously failed tests

stestr run also enables you to run just the tests that failed in the previous run. To do this you can use the ---failing argument.

A common workflow using this is:

# Run tests (and some fail)
$ stestr run
# Fix currently broken tests - repeat until there are no failures.
$ stestr run --failing
# Do a full run to find anything that regressed during the reduction process.
$ stestr run

The --failing option turns on --partial automatically (so that if the partial test run were to be interrupted, the failing tests that aren’t run are not lost).

Another common use case is repeating a failure that occurred on a remote machine (e.g. during a jenkins test run). There are a few common ways to do approach this.

Firstly, if you have a subunit stream from the run you can just load it:

$ stestr load < failing-stream
# Run the failed tests
$ stestr run --failing

The streams generated by test runs are in .stestr/ named for their test id - e.g. .stestr/0 is the first stream. Note for right now these files are stored in subunit v1, but all of stestr commands (including load) expect a subunit v2 stream.

If you have access to the remote machine you can also get the subunit stream by running:

$ stestr last --subunit > failing-stream

This is often a bit easier than trying to manually pull the stream file out of the .stestr directory. (also it will be in subunit v2)

If you do not have a stream or access to the machine you may be able to use a list file. If you can get a file that contains one test id per line, you can run the named tests like this:

$ stestr run --load-list FILENAME

This can also be useful when dealing with sporadically failing tests, or tests that only fail in combination with some other test - you can bisect the tests that were run to get smaller and smaller (or larger and larger) test subsets until the error is pinpointed.

stestr run --until-failure will run your test suite again and again and again stopping only when interrupted or a failure occurs. This is useful for repeating timing-related test failures.

Listing tests

To see a list of tests found by stestr you can use the stestr list command. This will list all tests found by discovery.

You can also use this to see what tests will be run by a given stestr run command. For instance, the tests that stestr run myfilter will run are shown by stestr list myfilter. As with the run command, arguments to list are used to regex filter the tests.

Parallel testing

stestr lets you run tests in parallel by default. So, it actually does this by def:

$ stestr run

This will first list the tests, partition the tests into one partition per CPU on the machine, and then invoke multiple test runners at the same time, with each test runner getting one partition. Currently the partitioning algorithm is simple round-robin for tests that stestr has not seen run before, and equal-time buckets for tests that stestr has seen run.

To determine how many CPUs are present in the machine, stestr will use the multiprocessing Python module On operating systems where this is not implemented, or if you need to control the number of workers that are used, the –concurrency option will let you do so:

$ stestr run --concurrency=2

When running tests in parallel, stestr adds a tag for each test to the subunit stream to show which worker executed that test. The tags are of the form worker-%d and are usually used to reproduce test isolation failures, where knowing exactly what test ran on a given worker is important. The %d that is substituted in is the partition number of tests from the test run - all tests in a single run with the same worker-N ran in the same test runner instance.

To find out which slave a failing test ran on just look at the ‘tags’ line in its test error:

======================================================================
label: testrepository.tests.ui.TestDemo.test_methodname
tags: foo worker-0
----------------------------------------------------------------------
error text

And then find tests with that tag:

$ stestr last --subunit | subunit-filter -s --xfail --with-tag=worker-3 | subunit-ls > slave-3.list

Grouping Tests

In certain scenarios you may want to group tests of a certain type together so that they will be run by the same worker process. The group_regex option in the stestr config file permits this. When set, tests are grouped by the group(0) of any regex match. Tests with no match are not grouped.

For example, setting the following option in the stestr config file will group tests in the same class together (the last . splits the class and test method):

group_regex=([^\.]+\.)+

Test Scheduling

By default stestr schedules the tests by first checking if there is any historical timing data on any tests. It then sorts the tests by that timing data loops over the tests in order and adds one to each worker that it will launch. For tests without timing data, the same is done, except the tests are in alphabetical order instead of based on timing data. If a group regex is used the same algorithm is used with groups instead of individual tests.

However there are options to adjust how stestr will schedule tests. The primary option to do this is to manually schedule all the tests run. To do this use the --worker-file option for stestr run. This takes a path to a yaml file that instructs stestr how to run tests. It is formatted as a list of dicts with a single element each with a list describing the tests to run on each worker. For example:

- worker:
  - regex 1

- worker:
  - regex 2
  - regex 3

would create 2 workers. The first would run all tests that match regex 1, and the second would run all tests that match regex 2 or regex 3.

There is also an option on stestr run, --random/-r to randomize the order of tests as they are passed to the workers. This is useful in certain use cases, especially when you want to test isolation between test cases.

Automated test isolation bisection

As mentioned above, its possible to manually analyze test isolation issues by interrogating the repository for which tests ran on which worker, and then creating a list file with those tests, re-running only half of them, checking the error still happens, rinse and repeat.

However that is tedious. stestr can perform this analysis for you:

$ stestr run --analyze-isolation

will perform that analysis for you. The process is:

1. The last run in the repository is used as a basis for analysing against - tests are only cross checked against tests run in the same worker in that run. This means that failures accrued from several different runs would not be processed with the right basis tests - you should do a full test run to seed your repository. This can be local, or just stestr load a full run from your Jenkins or other remote run environment.
2. Each test that is currently listed as a failure is run in a test process given just that id to run.
3. Tests that fail are excluded from analysis - they are broken on their own.
4. The remaining failures are then individually analysed one by one.
5. For each failing, it gets run in one work along with the first 1/2 of the tests that were previously run prior to it.
6. If the test now passes, that set of prior tests are discarded, and the other half of the tests is promoted to be the full list. If the test fails then other other half of the tests are discarded and the current set promoted.
7. Go back to running the failing test along with 1/2 of the current list of priors unless the list only has 1 test in it. If the failing test still failed with that test, we have found the isolation issue. If it did not then either the isolation issue is racy, or it is a 3-or-more test isolation issue. Neither of those cases are automated today.

Forcing isolation

Sometimes it is useful to force a separate test runner instance for each test executed. The --isolated flag will cause stestr to execute a separate runner per test:

$ stestr run --isolated

In this mode stestr first determines tests to run (either automatically listed, using the failing set, or a user supplied load-list), and then spawns one test runner per test it runs. To avoid cross-test-runner interactions concurrency is disabled in this mode. --analyze-isolation supersedes --isolated if they are both supplied.

Repositories

stestr uses a data repository to keep track of test previous test runs. There are different backend types that each offer different advantages. There are currently 2 repository types to choose from, file and sql.

You can choose which repository type you want with the --repo-type/-r cli flag. file is the current default.

You can also specify an alternative repository with the --repo-url/-u cli flags. The default value for a file repository type is to use the directory: $CWD/.stestr. For a sql repository type is to use a sqlite database located at: $CWD/.stestr.sqlite.

Note

Make sure you put these flags before the cli subcommand

Note

Different repository types that use local storage will conflict with each other in the same directory. If you initialize one repository type and then try to use another in the same directory, it will not work.

File

The default stestr repository type has a very simple disk structure. It contains the following files:

• format: This file identifies the precise layout of the repository, in case future changes are needed.
• next-stream: This file contains the serial number to be used when adding another stream to the repository.
• failing: This file is a stream containing just the known failing tests. It is updated whenever a new stream is added to the repository, so that it only references known failing tests.
• #N - all the streams inserted in the repository are given a serial number.

SQL

This is an experimental repository backend, that is based on the subunit2sql library. It’s currently still under development and should be considered experimental for the time being. Eventually it’ll replace the File repository type

Note

The sql repository type requirements are not installed by default. They are listed under the ‘sql’ setuptools extras. You can install them with pip by running: pip install 'stestr[sql]'