freeipa/ipatests/pytest_ipa/slicing.py

201 lines
6.3 KiB
Python
Raw Normal View History

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 <slev@altlinux.org> Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>
2019-07-03 04:56:32 -05:00
#
# 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()