From a2d4e2a61f46cf6337786b2cb031e3a9f394a708 Mon Sep 17 00:00:00 2001 From: Stanislav Levin Date: Wed, 3 Jul 2019 12:56:32 +0300 Subject: [PATCH] Make use of Azure Pipeline slicing The unit tests execution time within Azure Pipelines(AP) is not balanced. One test job(Base) takes ~13min, while another(XMLRPC) ~28min. Fortunately, AP supports slicing: > An agent job can be used to run a suite of tests in parallel. For example, you can run a large suite of 1000 tests on a single agent. Or, you can use two agents and run 500 tests on each one in parallel. To leverage slicing, the tasks in the job should be smart enough to understand the slice they belong to. >The step that runs the tests in a job needs to know which test slice should be run. The variables System.JobPositionInPhase and System.TotalJobsInPhase can be used for this purpose. Thus, to support this pytest should know how to split the test suite into groups(slices). For this, a new internal pytest plugin was added. About plugin. - Tests within a slice are grouped by test modules because not all of the tests within the module are independent from each other. - Slices are balanced by the number of tests within test module. - To run some module within its own environment there is a dedicated slice option (could help with extremely slow tests) Examples. - To split `test_cmdline` tests into 2 slices and run the first one: ipa-run-tests --slices=2 --slice-num=1 test_cmdline - To split tests into 2 slices, then to move one module out to its own slice and run the second one: ipa-run-tests --slices=2 --slice-dedicated=test_cmdline/test_cli.py \ --slice-num=2 test_cmdline Fixes: https://pagure.io/freeipa/issue/8008 Signed-off-by: Stanislav Levin Reviewed-By: Alexander Bokovoy --- ipatests/azure/azure-pipelines.yml | 18 +- ipatests/azure/azure-run-tests.sh | 9 +- ipatests/azure/templates/run-test.yml | 4 + ipatests/azure/templates/test-jobs.yml | 7 +- ipatests/conftest.py | 1 + ipatests/pytest_ipa/slicing.py | 200 ++++++++++++++++++ .../test_ipatests_plugins/test_slicing.py | 127 +++++++++++ 7 files changed, 351 insertions(+), 15 deletions(-) create mode 100644 ipatests/pytest_ipa/slicing.py create mode 100644 ipatests/test_ipatests_plugins/test_slicing.py diff --git a/ipatests/azure/azure-pipelines.yml b/ipatests/azure/azure-pipelines.yml index 89f81b87f..6d75956bd 100644 --- a/ipatests/azure/azure-pipelines.yml +++ b/ipatests/azure/azure-pipelines.yml @@ -128,8 +128,8 @@ jobs: - template: templates/test-jobs.yml parameters: - jobName: Base - jobTitle: Base tests + jobName: BASE_XMLRPC + jobTitle: BASE and XMLRPC tests testsToRun: - test_cmdline - test_install @@ -138,20 +138,12 @@ jobs: - test_ipaplatform - test_ipapython - test_ipatests_plugins - testsToIgnore: - - test_integration - - test_webui - - test_ipapython/test_keyring.py - taskToRun: run-tests - -- template: templates/test-jobs.yml - parameters: - jobName: XMLRPC - jobTitle: XMLRPC tests - testsToRun: - test_xmlrpc testsToIgnore: - test_integration - test_webui - test_ipapython/test_keyring.py + testsToDedicate: + - test_xmlrpc/test_dns_plugin.py taskToRun: run-tests + tasksParallel: 3 diff --git a/ipatests/azure/azure-run-tests.sh b/ipatests/azure/azure-run-tests.sh index aef13568e..c66e59028 100755 --- a/ipatests/azure/azure-run-tests.sh +++ b/ipatests/azure/azure-run-tests.sh @@ -6,6 +6,9 @@ server_password=Secret123 # Normalize spacing and expand the list afterwards. Remove {} for the single list element case tests_to_run=$(eval "echo {$(echo $TESTS_TO_RUN | sed -e 's/[ \t]+*/,/g')}" | tr -d '{}') tests_to_ignore=$(eval "echo --ignore\ {$(echo $TESTS_TO_IGNORE | sed -e 's/[ \t]+*/,/g')}" | tr -d '{}') +tests_to_dedicate= +[[ -n "$TESTS_TO_DEDICATE" ]] && \ +tests_to_dedicate=$(eval "echo --slice-dedicated={$(echo $TESTS_TO_DEDICATE | sed -e 's/[ \t]+*/,/g')}" | tr -d '{}') systemctl --now enable firewalld echo "Installing FreeIPA master for the domain ${server_domain} and realm ${server_realm}" @@ -39,7 +42,11 @@ if [ "$install_result" -eq 0 ] ; then ipa-test-task --help ipa-run-tests --help - ipa-run-tests ${tests_to_ignore} --verbose --with-xunit '-k not test_dns_soa' ${tests_to_run} + ipa-run-tests ${tests_to_ignore} \ + ${tests_to_dedicate} \ + --slices=${SYSTEM_TOTALJOBSINPHASE:-1} \ + --slice-num=${SYSTEM_JOBPOSITIONINPHASE:-1} \ + --verbose --with-xunit '-k not test_dns_soa' ${tests_to_run} tests_result=$? else echo "ipa-server-install failed with code ${save_result}, skip IPA tests" diff --git a/ipatests/azure/templates/run-test.yml b/ipatests/azure/templates/run-test.yml index 1e4498419..301115f16 100644 --- a/ipatests/azure/templates/run-test.yml +++ b/ipatests/azure/templates/run-test.yml @@ -5,6 +5,7 @@ parameters: taskToRun: 'run-tests' testsToRun: '' testsToIgnore: '' + testsToDedicate: '' steps: - script: | @@ -22,7 +23,10 @@ steps: set -e docker exec --env TESTS_TO_RUN="${{ parameters.testsToRun }}" \ --env TESTS_TO_IGNORE="${{ parameters.testsToIgnore }}" \ + --env TESTS_TO_DEDICATE="${{ parameters.testsToDedicate }}" \ --env CI_RUNNER_LOGS_DIR="${{ parameters.logsPath }}" \ + --env SYSTEM_TOTALJOBSINPHASE=$(System.TotalJobsInPhase) \ + --env SYSTEM_JOBPOSITIONINPHASE=$(System.JobPositionInPhase) \ --privileged -t \ $(createContainer.containerName) \ /bin/bash --noprofile --norc -x /freeipa/ipatests/azure/azure-${{parameters.taskToRun}}.sh diff --git a/ipatests/azure/templates/test-jobs.yml b/ipatests/azure/templates/test-jobs.yml index c79058c9e..5b0173fdf 100644 --- a/ipatests/azure/templates/test-jobs.yml +++ b/ipatests/azure/templates/test-jobs.yml @@ -3,7 +3,9 @@ parameters: jobTitle: '' testsToIgnore: [] testsToRun: [] + testsToDedicate: [] taskToRun: '' + tasksParallel: 1 jobs: - job: ${{ parameters.jobName }} @@ -12,6 +14,8 @@ jobs: condition: succeeded() pool: vmImage: 'Ubuntu-16.04' + strategy: + parallel: ${{ parameters.tasksParallel }} steps: - template: setup-test-environment.yml - template: run-test.yml @@ -21,6 +25,7 @@ jobs: taskToRun: ${{ parameters.taskToRun}} testsToRun: ${{ join(' ', parameters.testsToRun ) }} testsToIgnore: ${{ join(' ', parameters.testsToIgnore ) }} + testsToDedicate: ${{ join(' ', parameters.testsToDedicate ) }} - task: PublishTestResults@2 inputs: testResultsFiles: $(CI_RUNNER_LOGS_DIR)/nosetests.xml @@ -28,4 +33,4 @@ jobs: condition: succeededOrFailed() - template: save-test-artifacts.yml parameters: - logsArtifact: logs-${{parameters.jobName}}-$(Build.BuildId)-$(Agent.OS)-$(Agent.OSArchitecture) + logsArtifact: logs-${{parameters.jobName}}-$(Build.BuildId)-$(System.JobPositionInPhase)-$(Agent.OS)-$(Agent.OSArchitecture) diff --git a/ipatests/conftest.py b/ipatests/conftest.py index ae523c88a..62045de31 100644 --- a/ipatests/conftest.py +++ b/ipatests/conftest.py @@ -25,6 +25,7 @@ HERE = os.path.dirname(os.path.abspath(__file__)) pytest_plugins = [ 'ipatests.pytest_ipa.additional_config', + 'ipatests.pytest_ipa.slicing', 'ipatests.pytest_ipa.beakerlib', 'ipatests.pytest_ipa.declarative', 'ipatests.pytest_ipa.nose_compat', diff --git a/ipatests/pytest_ipa/slicing.py b/ipatests/pytest_ipa/slicing.py new file mode 100644 index 000000000..172208514 --- /dev/null +++ b/ipatests/pytest_ipa/slicing.py @@ -0,0 +1,200 @@ +# +# Copyright (C) 2019 FreeIPA Contributors see COPYING for license +# + +""" +The main purpose of this plugin is to slice a test suite into +several pieces to run each within its own test environment(for example, +an Agent of Azure Pipelines). + +Tests within a slice are grouped by test modules because not all of the tests +within the module are independent from each other. + +Slices are balanced by the number of tests within test module. +* Actually, tests should be grouped by the execution duration. +This could be achieved by the caching of tests results. Azure Pipelines +caching is in development. * +To workaround slow tests a dedicated slice is added. + +:param slices: A total number of slices to split the test suite into +:param slice-num: A number of slice to run +:param slice-dedicated: A file path to the module to run in its own slice + +**Examples** + +Inputs: +ipa-run-tests test_cmdline --collectonly -qq +... +test_cmdline/test_cli.py: 39 +test_cmdline/test_help.py: 7 +test_cmdline/test_ipagetkeytab.py: 16 +... + +* Split tests into 2 slices and run the first one: + +ipa-run-tests --slices=2 --slice-num=1 test_cmdline + +The outcome would be: +... +Running slice: 1 (46 tests) +Modules: +test_cmdline/test_cli.py: 39 +test_cmdline/test_help.py: 7 +... + +* Split tests into 2 slices, move one module out to its own slice +and run the second one + +ipa-run-tests --slices=2 --slice-dedicated=test_cmdline/test_cli.py \ + --slice-num=2 test_cmdline + +The outcome would be: +... +Running slice: 2 (23 tests) +Modules: +test_cmdline/test_ipagetkeytab.py: 16 +test_cmdline/test_help.py: 7 +... + +""" +import pytest + + +def pytest_addoption(parser): + group = parser.getgroup("slicing") + group.addoption( + '--slices', dest='slices_num', type=int, + help='The number of slices to split the test suite into') + group.addoption( + '--slice-num', dest='slice_num', type=int, + help='The specific number of slice to run') + group.addoption( + '--slice-dedicated', action="append", dest='slices_dedicated', + help='The file path to the module to run in dedicated slice') + + +@pytest.hookimpl(hookwrapper=True) +def pytest_collection_modifyitems(session, config, items): + yield + slice_count = config.getoption('slices_num') + slice_id = config.getoption('slice_num') + modules_dedicated = config.getoption('slices_dedicated') + # deduplicate + if modules_dedicated: + modules_dedicated = list(set(modules_dedicated)) + + # sanity check + if not slice_count or not slice_id: + return + + # nothing to do + if slice_count == 1: + return + + if modules_dedicated and len(modules_dedicated) > slice_count: + raise ValueError( + "Dedicated slice number({}) shouldn't be greater than the number " + "of slices({})".format(len(modules_dedicated), slice_count)) + + if slice_id > slice_count: + raise ValueError( + "Slice number({}) shouldn't be greater than the number of slices" + "({})".format(slice_id, slice_count)) + + modules = [] + # Calculate modules within collection + # Note: modules within pytest collection could be placed in not consecutive + # order + for number, item in enumerate(items): + name = item.nodeid.split("::", 1)[0] + if not modules or name != modules[-1]["name"]: + modules.append({"name": name, "begin": number, "end": number}) + else: + modules[-1]["end"] = number + + if slice_count > len(modules): + raise ValueError( + "Total number of slices({}) shouldn't be greater than the number " + "of Python test modules({})".format(slice_count, len(modules))) + + slices_dedicated = [] + if modules_dedicated: + slices_dedicated = [ + [m] for m in modules for x in modules_dedicated if x in m["name"] + ] + if modules_dedicated and len(slices_dedicated) != len(modules_dedicated): + raise ValueError( + "The number of dedicated slices({}) should be equal to the " + "number of dedicated modules({})".format( + slices_dedicated, modules_dedicated)) + + if (slices_dedicated and len(slices_dedicated) == slice_count and + len(slices_dedicated) != len(modules)): + raise ValueError( + "The total number of slices({}) is not sufficient to run dedicated" + " modules({}) as well as usual ones({})".format( + slice_count, len(slices_dedicated), + len(modules) - len(slices_dedicated))) + + # remove dedicated modules from usual ones + for s in slices_dedicated: + for m in s: + if m in modules: + modules.remove(m) + + avail_slice_count = slice_count - len(slices_dedicated) + # initialize slices with empty lists + slices = [[] for i in range(slice_count)] + + # initialize slices with dedicated ones + for sn, s in enumerate(slices_dedicated): + slices[sn] = s + + # initial reverse sort by the number of tests in a test module + modules.sort(reverse=True, key=lambda x: x["end"] - x["begin"] + 1) + reverse = True + while modules: + for sslice_num, sslice in enumerate(sorted( + modules[:avail_slice_count], + reverse=reverse, key=lambda x: x["end"] - x["begin"] + 1)): + slices[len(slices_dedicated) + sslice_num].append(sslice) + + modules[:avail_slice_count] = [] + reverse = not reverse + + calc_ntests = sum(x["end"] - x["begin"] + 1 for s in slices for x in s) + assert calc_ntests == len(items) + assert len(slices) == slice_count + + # the range of the given argument `slice_id` begins with 1(one) + sslice = slices[slice_id - 1] + + new_items = [] + for m in sslice: + new_items += items[m["begin"]:m["end"] + 1] + items[:] = new_items + + tw = config.get_terminal_writer() + if tw: + tw.line() + tw.write( + "Running slice: {} ({} tests)\n".format( + slice_id, + len(items), + ), + cyan=True, + bold=True, + ) + tw.write( + "Modules:\n", + yellow=True, + bold=True, + ) + for module in sslice: + tw.write( + "{}: {}\n".format( + module["name"], + module["end"] - module["begin"] + 1), + yellow=True, + ) + tw.line() diff --git a/ipatests/test_ipatests_plugins/test_slicing.py b/ipatests/test_ipatests_plugins/test_slicing.py new file mode 100644 index 000000000..e21af33a1 --- /dev/null +++ b/ipatests/test_ipatests_plugins/test_slicing.py @@ -0,0 +1,127 @@ +# +# Copyright (C) 2019 FreeIPA Contributors see COPYING for license +# + +import glob + +import pytest + +MOD_NAME = "test_module_{}" +FUNC_NAME = "test_func_{}" +PYTEST_INTERNAL_ERROR = 3 +MODS_NUM = 5 + + +@pytest.fixture +def ipatestdir(testdir): + """ + Create MODS_NUM test modules within testdir. + Each module contains 1 test function. + """ + testdir.makeconftest( + """ + pytest_plugins = ["ipatests.pytest_ipa.slicing"] + """ + ) + for i in range(MODS_NUM): + testdir.makepyfile( + **{MOD_NAME.format(i): + """ + def {func}(): + pass + """.format(func=FUNC_NAME.format(i)) + } + ) + return testdir + + +@pytest.mark.parametrize( + "nslices,nslices_d,groups", + [(2, 0, [[x for x in range(MODS_NUM) if x % 2 == 0], + [x for x in range(MODS_NUM) if x % 2 != 0]]), + (2, 1, [[0], [x for x in range(1, MODS_NUM)]]), + (1, 0, [[x for x in range(MODS_NUM)]]), + (1, 1, [[x for x in range(MODS_NUM)]]), + (MODS_NUM, MODS_NUM, [[x] for x in range(MODS_NUM)]), + ]) +def test_slicing(ipatestdir, nslices, nslices_d, groups): + """ + Positive tests. + + Run `nslices` slices, including `nslices_d` dedicated slices. + The `groups` is an expected result of slices grouping. + + For example, there are 5 test modules. If one runs them in + two slices (without dedicated ones) the expected result will + be [[0, 2, 4], [1, 3]]. This means, that first slice will run + modules 0, 2, 4, second one - 1 and 3. + + Another example, there are 5 test modules. We want to run them + in two slices. Also we specify module 0 as dedicated. + The expected result will be [[0], [1, 2, 3, 4]], which means, that + first slice will run module 0, second one - 1, 2, 3, 4. + + If the given slice count is one, then this plugin does nothing. + """ + for sl in range(nslices): + args = [ + "-v", + "--slices={}".format(nslices), + "--slice-num={}".format(sl + 1) + ] + for dslice in range(nslices_d): + args.append( + "--slice-dedicated={}.py".format(MOD_NAME.format(dslice))) + result = ipatestdir.runpytest(*args) + assert result.ret == 0 + result.assert_outcomes(passed=len(groups[sl])) + for mod_num in groups[sl]: + result.stdout.fnmatch_lines(["*{mod}.py::{func} PASSED*".format( + mod=MOD_NAME.format(mod_num), + func=FUNC_NAME.format(mod_num))]) + + +@pytest.mark.parametrize( + "nslices,nslices_d,nslice,dmod,err_message", + [(2, 3, 1, None, + "Dedicated slice number({}) shouldn't be greater than" + " the number of slices({})".format(3, 2)), + (MODS_NUM, 0, MODS_NUM + 1, None, + "Slice number({}) shouldn't be greater than the number of slices" + "({})".format( + MODS_NUM + 1, MODS_NUM)), + (MODS_NUM + 1, 1, 1, None, + "Total number of slices({}) shouldn't be greater" + " than the number of Python test modules({})".format( + MODS_NUM + 1, MODS_NUM)), + (MODS_NUM, MODS_NUM, 1, "notexisted_module", + "The number of dedicated slices({}) should be equal to the " + "number of dedicated modules({})".format( + [], ["notexisted_module.py"])), + (MODS_NUM - 1, MODS_NUM - 1, 1, None, + "The total number of slices({}) is not sufficient to" + " run dedicated modules({}) as well as usual ones({})".format( + MODS_NUM - 1, MODS_NUM - 1, 1)), + ]) +def test_slicing_negative(ipatestdir, nslices, nslices_d, nslice, dmod, + err_message): + """ + Negative scenarios + """ + args = [ + "-v", + "--slices={}".format(nslices), + "--slice-num={}".format(nslice) + ] + if dmod is None: + for dslice in range(nslices_d): + args.append( + "--slice-dedicated={}.py".format(MOD_NAME.format(dslice))) + else: + args.append( + "--slice-dedicated={}.py".format(dmod)) + result = ipatestdir.runpytest(*args) + assert result.ret == PYTEST_INTERNAL_ERROR + result.assert_outcomes() + result.stdout.fnmatch_lines(["*ValueError: {err_message}*".format( + err_message=glob.escape(err_message))])