Arvados

h1. Running tests

{{toc}}

The arvados git repository has a @run-tests.sh@ script which tests (nearly) all of the components in the source tree. Jenkins at https://ci.arvados.org uses this exact script, so running it _before pushing a new master_ is a good way to predict whether Jenkins will fail your build and start pestering you by IRC.

h2. Background

You have the arvados source tree at @~/arvados@ and you might have local modifications.

h2. Install prerequisites

Follow instructions at [[Hacking prerequisites]]: "Install dev environment", etc. Don't miss creating the Postgres database, docker groups, etc.

h2. Environment

Your locale must use utf-8. Set environment variable LANG=C.UTF-8 if necessary to ensure the "locale" command reports UTF-8.

h2. If you're Jenkins

Run all tests from a clean slate (slow, but more immune to leaks)

<pre>

~/arvados/build/run-tests.sh WORKSPACE=~/arvados

</pre>

h2. If you're a developer

Cache everything needed to run test suites:

<pre>

mkdir ~/.cache/arvados-build

~/arvados/build/run-tests.sh WORKSPACE=~/arvados --temp ~/.cache/arvados-build --only install

</pre>

Start interactive mode:

<pre>

$ ~/arvados/build/run-tests.sh WORKSPACE=~/arvados --temp ~/.cache/arvados-build --interactive

</pre>

Start interactive mode and enabled debug output:

<pre>

$ ~/arvados/build/run-tests.sh WORKSPACE=~/arvados ARVADOS_DEBUG=1 --temp ~/.cache/arvados-build --interactive

</pre>

When prompted, choose a test suite to run:

<pre>

== Interactive commands:

test TARGET

test TARGET:py3        (test with python3)

test TARGET -check.vv  (pass arguments to test)

install TARGET

install env            (go/python libs)

install deps           (go/python libs + arvados components needed for integration tests)

reset                  (...services used by integration tests)

exit

== Test targets:

cmd/arvados-client              lib/dispatchcloud/container     sdk/go/auth                     sdk/pam:py3                     services/fuse                   tools/crunchstat-summary

cmd/arvados-server              lib/dispatchcloud/scheduler     sdk/go/blockdigest              sdk/python                      services/fuse:py3               tools/crunchstat-summary:py3

lib/cli                         lib/dispatchcloud/ssh_executor  sdk/go/crunchrunner             sdk/python:py3                  services/health                 tools/keep-block-check

lib/cloud                       lib/dispatchcloud/worker        sdk/go/dispatch                 services/arv-git-httpd          services/keep-balance           tools/keep-exercise

lib/cloud/azure                 lib/service                     sdk/go/health                   services/crunch-dispatch-local  services/keepproxy              tools/keep-rsync

lib/cloud/ec2                   sdk/cwl                         sdk/go/httpserver               services/crunch-dispatch-slurm  services/keepstore              tools/sync-groups

lib/cmd                         sdk/cwl:py3                     sdk/go/keepclient               services/crunch-run             services/keep-web

lib/controller                  sdk/go/arvados                  sdk/go/manifest                 services/crunchstat             services/nodemanager

lib/crunchstat                  sdk/go/arvadosclient            sdk/go/stats                    services/dockercleaner          services/nodemanager:py3

lib/dispatchcloud               sdk/go/asyncbuf                 sdk/pam                         services/dockercleaner:py3      services/ws

What next? 

</pre>

Example: testing lib/dispatchcloud/container, showing verbose/debug logs:

<pre>

What next? test lib/dispatchcloud/container/ -check.vv

======= test lib/dispatchcloud/container

START: queue_test.go:99: IntegrationSuite.TestCancelIfNoInstanceType

WARN[0000] cancel container with no suitable instance type  ContainerUUID=zzzzz-dz642-queuedcontainer error="no suitable instance type"

WARN[0000] cancel container with no suitable instance type  ContainerUUID=zzzzz-dz642-queuedcontainer error="no suitable instance type"

START: queue_test.go:37: IntegrationSuite.TearDownTest

PASS: queue_test.go:37: IntegrationSuite.TearDownTest   0.846s

PASS: queue_test.go:99: IntegrationSuite.TestCancelIfNoInstanceType     0.223s

START: queue_test.go:42: IntegrationSuite.TestGetLockUnlockCancel

INFO[0001] adding container to queue                     ContainerUUID=zzzzz-dz642-queuedcontainer InstanceType=testType Priority=1 State=Queued

START: queue_test.go:37: IntegrationSuite.TearDownTest

PASS: queue_test.go:37: IntegrationSuite.TearDownTest   0.901s

PASS: queue_test.go:42: IntegrationSuite.TestGetLockUnlockCancel        0.177s

OK: 2 passed

PASS

ok      git.arvados.org/arvados.git/lib/dispatchcloud/container       2.150s

======= test lib/dispatchcloud/container -- 3s

Pass: lib/dispatchcloud/container tests (3s)

All test suites passed.

</pre>

h3. Running individual test cases

h4. Golang

Most Go packages use gocheck. Use gocheck command line args like -check.f.

<pre>

What next? test lib/dispatchcloud/container -check.vv -check.f=LockUnlock

======= test lib/dispatchcloud/container

START: queue_test.go:42: IntegrationSuite.TestGetLockUnlockCancel

INFO[0000] adding container to queue                     ContainerUUID=zzzzz-dz642-queuedcontainer InstanceType=testType Priority=1 State=Queued

START: queue_test.go:37: IntegrationSuite.TearDownTest

PASS: queue_test.go:37: IntegrationSuite.TearDownTest   0.812s

PASS: queue_test.go:42: IntegrationSuite.TestGetLockUnlockCancel        0.184s

OK: 1 passed

PASS

ok      git.arvados.org/arvados.git/lib/dispatchcloud/container       1.000s

======= test lib/dispatchcloud/container -- 2s

</pre>

h4. Python

If what you really want to do is focus on failing or newly-added tests, consider passing the appropriate switches to do that:

<pre>

  -x, --exitfirst       Exit instantly on first error or failed test

  --lf, --last-failed   Rerun only the tests that failed at the last run (or

                        all if none failed)

  --ff, --failed-first  Run all tests, but run the last failures first. This

                        may re-order tests and thus lead to repeated fixture

                        setup/teardown.

  --nf, --new-first     Run tests from new files first, then the rest of the

                        tests sorted by file mtime

</pre>

If you want to manually select tests:

<pre>

  FILENAME              Run tests from FILENAME, relative to the source root

  FILENAME::CLASSNAME   Run tests from CLASSNAME

  FILENAME::FUNCNAME, FILENAME::CLASSNAME::FUNCNAME

                        Run only the named test function

  -k EXPRESSION         Only run tests which match the given substring

                        expression. An expression is a Python evaluable

                        expression where all names are substring-matched

                        against test names and their parent classes.

                        Example: -k 'test_method or test_other' matches all

                        test functions and classes whose name contains

                        'test_method' or 'test_other', while -k 'not

                        test_method' matches those that don't contain

                        'test_method' in their names. -k 'not test_method

                        and not test_other' will eliminate the matches.

                        Additionally keywords are matched to classes and

                        functions containing extra names in their

                        'extra_keyword_matches' set, as well as functions

                        which have names assigned directly to them. The

                        matching is case-insensitive.

  -m MARKEXPR           Only run tests matching given mark expression. For

                        example: -m 'mark1 and not mark2'.

</pre>

For even more options, refer to the "pytest command line reference":https://docs.pytest.org/en/stable/reference/reference.html#command-line-flags.

Example:

<pre>

What next? test sdk/python:py3 --disable-warnings --tb=no --no-showlocals tests/test_keep_client.py::KeepDiskCacheTestCase

======= test sdk/python

[…pip output…]

========================================================== test session starts ==========================================================

platform linux -- Python 3.8.19, pytest-8.2.0, pluggy-1.5.0

rootdir: /home/brett/Curii/arvados/sdk/python

configfile: pytest.ini

collected 9 items

tests/test_keep_client.py F........                                                                                               [100%]

======================================================== short test summary info ========================================================

FAILED tests/test_keep_client.py::KeepDiskCacheTestCase::test_disk_cache_cap - AssertionError: True is not false

====================================================== 1 failed, 8 passed in 0.16s ======================================================

======= sdk/python tests -- FAILED

======= test sdk/python -- 2s

Failures (1):

Fail: sdk/python tests (2s)

What next? test sdk/python:py3 --disable-warnings --tb=no --no-showlocals --lf

======= test sdk/python

[…pip output…]

========================================================== test session starts ==========================================================

platform linux -- Python 3.8.19, pytest-8.2.0, pluggy-1.5.0

rootdir: /home/brett/Curii/arvados/sdk/python

configfile: pytest.ini

testpaths: tests

collected 964 items / 963 deselected / 1 selected

run-last-failure: rerun previous 1 failure

tests/test_keep_client.py F                                                                                                       [100%]

======================================================== short test summary info ========================================================

FAILED tests/test_keep_client.py::KeepDiskCacheTestCase::test_disk_cache_cap - AssertionError: True is not false

============================================= 1 failed, 963 deselected, 1 warning in 0.43s ==============================================

======= sdk/python tests -- FAILED

======= test sdk/python -- 2s

Failures (1):

Fail: sdk/python tests (2s)

</pre>

h4. RailsAPI

<pre>

What next? test services/api TESTOPTS=--name=/.*signed.locators.*/

[...]

# Running:

....

Finished in 1.080084s, 3.7034 runs/s, 461.0751 assertions/s.

</pre>

h3. Restarting services for integration tests

If you have changed services/api code, and you want to check whether it breaks the lib/dispatchcloud/container integration tests:

<pre>

What next? reset                                # teardown the integration-testing environment

What next? install services/api                 # (only needed if you've updated dependencies)

What next? test lib/dispatchcloud/container     # bring up the integration-testing environment and run tests

What next? test lib/dispatchcloud/container     # leave the integration-testing environment up and run tests

</pre>

h3. Updating cache after pulling master

Always quit interactive mode and restart after modifying run-tests.sh (via git-pull, git-checkout, editing, etc).

When you start, run "install all" to get the latest gem/python dependencies, install updated versions of Arvados services used by integration tests, etc.

Then you can resume your cycle of "test lib/controller", etc.

h3. Controlling test order (Rails)

Rails tests start off with a line like this

<pre>

Run options: -v -d --seed 57089

</pre>

The seed value determines the order tests are run. To reproduce reproduce an order-dependent test failure, specify the same seed as a previous failed run:

<pre>

What next? test services/api TESTOPTS="-v -d --seed 57089"

</pre>

h3. Other options

For more usage info, try:

<pre>

~/arvados/build/run-tests.sh --help

</pre>

h2. Running workbench diagnostics tests

You can run workbench diagnostics tests against any production server. 

Update your workbench application.yml to add a "diagnostics" section with the login token and pipeline details. The below example configuration is to run the "qr1hi-p5p6p-ftcb0o61u4yd2zr" pipeline in "qr1hi" environment.

<pre>

diagnostics:

  secret_token: useanicelongrandomlygeneratedsecrettokenstring

  arvados_workbench_url: https://workbench.qr1hi.arvadosapi.com

  user_tokens:

    active: yourcurrenttokenintheenvironmenttowhichyouarepointing

  pipelines_to_test:

    pipeline_1:

      template_uuid: qr1hi-p5p6p-ftcb0o61u4yd2zr

      input_paths: []

      max_wait_seconds: 300

</pre>

You can now run the "qr1hi" diagnostics tests using the following command:

<pre>

  cd $ARVADOS_HOME

  RAILS_ENV=diagnostics bundle exec rake TEST=test/diagnostics/pipeline_test.rb

</pre>

h2. Running workbench2 tests

React uses a lot of filesystem watchers (via inotify). The default number of watched files is relatively low at 8192. Increase that with:

  echo fs.inotify.max_user_watches=524288 | sudo tee -a /etc/sysctl.conf && sudo sysctl -p

h3. Docker

The integration tests can be run on non-debian based systems using docker. The workbench2 repo includes a Dockerfile that preinstalls the necessary dependencies to run cypress.

Build the docker image using this command from within the @arvados@ repository

<pre>

docker build . -f services/workbench2/docker/Dockerfile -t workbench2-build

</pre>

Then, start the container using the following arguments:

<pre>

xhost +local:root

ARVADOS_DIR=/path/to/arvados

docker run -ti -v$ARVADOS_DIR:/usr/src/arvados -w /usr/src/arvados --env="DISPLAY" --volume="/tmp/.X11-unix:/tmp/.X11-unix:rw" workbench2-build:latest bash

</pre>

h4. Docker Troubleshooting

h5. No version of Cypress is installed

Solution: Manually install the required cypress version by running @./node_modules/.bin/cypress install@ inside the container

h3. Debian Host System

These instructions assume a Debian 10 (buster) host system. 

Install the Arvados test dependencies:

<pre>

echo "deb http://deb.debian.org/debian buster-backports main" > /etc/apt/sources.list.d/backports.list

apt-get update

apt-get install -y --no-install-recommends golang -t buster-backports

apt-get install -y --no-install-recommends build-essential ca-certificates git libpam0g-dev

</pre>

Install a few more dependencies for workbench2:

<pre>

apt-get update

apt-get install -y --no-install-recommends gnupg2 sudo curl

curl -sS https://dl.yarnpkg.com/debian/pubkey.gpg | sudo apt-key add -

echo "deb https://dl.yarnpkg.com/debian/ stable main" | sudo tee /etc/apt/sources.list.d/yarn.list

apt-get update

apt-get install -y --no-install-recommends yarn libgbm-dev

# we need the correct version of node to install cypress

# use arvados-server install to get it (and all other dependencies)

# All this will then not need to be repeated by ./tools/run-integration-tests.sh

# so we are not doing double work.

cd /usr/src/arvados

go mod download

cd cmd/arvados-server

go install

~/go/bin/arvados-server install -type test

cd <your-wb2-directory>

yarn run cypress install

</pre>

Make sure you have both the arvados and arvados-workbench2 source trees available, and then use the following commands (adjust path for the arvados source tree, if necessary) from your workbench2 source tree.

h3. Running Tests

Run the unit tests with:

<pre>

make unit-tests

# or

yarn test

</pre>

Run the cypress integration tests with:

<pre>

./tools/run-integration-tests.sh -i -a /usr/src/arvados

</pre>