discourse/lib/turbo_tests/runner.rb

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

326 lines
9.0 KiB
Ruby
Raw Permalink Normal View History

2019-06-20 21:33:41 -04:00
# frozen_string_literal: true
module TurboTests
class Runner
def self.run(opts = {})
files = opts[:files]
formatters = opts[:formatters]
seed = opts[:seed]
start_time = opts.fetch(:start_time) { Time.now }
verbose = opts.fetch(:verbose, false)
fail_fast = opts.fetch(:fail_fast, nil)
use_runtime_info = opts.fetch(:use_runtime_info, false)
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
retry_and_log_flaky_tests = opts.fetch(:retry_and_log_flaky_tests, false)
STDOUT.puts "VERBOSE" if verbose
reporter =
Reporter.from_config(
formatters,
start_time,
max_timings_count: opts[:profile_print_slowest_examples_count],
)
if ENV["GITHUB_ACTIONS"]
RSpec.configure do |config|
# Enable color output in GitHub Actions
# This eventually will be `config.color_mode = :on` in RSpec 4?
config.tty = true
config.color = true
end
end
new(
reporter: reporter,
files: files,
verbose: verbose,
fail_fast: fail_fast,
use_runtime_info: use_runtime_info,
seed: seed,
profile: opts[:profile],
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
retry_and_log_flaky_tests: retry_and_log_flaky_tests,
).run
end
def self.default_spec_folders
# We do not want to include system specs by default, they are quite slow.
Dir
.entries("#{Rails.root}/spec")
.reject { |entry| !File.directory?("spec/#{entry}") || %w[.. . system].include?(entry) }
.map { |entry| "spec/#{entry}" }
end
def initialize(opts)
@reporter = opts[:reporter]
@files = opts[:files]
@verbose = opts[:verbose]
@fail_fast = opts[:fail_fast]
@use_runtime_info = opts[:use_runtime_info]
@seed = opts[:seed]
@profile = opts[:profile]
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
@retry_and_log_flaky_tests = opts[:retry_and_log_flaky_tests]
@failure_count = 0
@messages = Queue.new
@threads = []
@error = false
end
def run
check_for_migrations
@num_processes = ParallelTests.determine_number_of_processes(nil)
group_opts = {}
group_opts[:runtime_log] = "tmp/turbo_rspec_runtime.log" if @use_runtime_info
tests_in_groups =
ParallelTests::RSpec::Runner.tests_in_groups(@files, @num_processes, **group_opts)
setup_tmp_dir
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
@reporter.add_formatter(Flaky::FailuresLoggerFormatter.new) if @retry_and_log_flaky_tests
subprocess_opts = { record_runtime: @use_runtime_info }
start_multisite_subprocess(@files, **subprocess_opts)
tests_in_groups.each_with_index do |tests, process_id|
start_regular_subprocess(tests, process_id + 1, **subprocess_opts)
end
@reporter.start
handle_messages
@reporter.finish
@threads.each(&:join)
2019-07-09 03:51:23 -04:00
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
if @retry_and_log_flaky_tests && @reporter.failed_examples.present?
retry_failed_examples_threshold = 10
if @reporter.failed_examples.length <= retry_failed_examples_threshold
STDOUT.puts "Retrying failed examples and logging flaky tests..."
return rerun_failed_examples(@reporter.failed_examples)
else
STDOUT.puts "Retry and log flaky tests was enabled but ignored because there are more than #{retry_failed_examples_threshold} failures."
Flaky::Manager.remove_flaky_tests
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
end
end
@reporter.failed_examples.empty? && !@error
end
protected
def check_for_migrations
config =
ActiveRecord::Base
.configurations
.find_db_config("test")
.configuration_hash
.merge("database" => "discourse_test_1")
ActiveRecord::Tasks::DatabaseTasks.migrations_paths = %w[db/migrate db/post_migrate]
begin
ActiveRecord::Migration.check_all_pending!
rescue ActiveRecord::PendingMigrationError
puts "There are pending migrations, run rake parallel:migrate"
exit 1
end
end
def setup_tmp_dir
begin
FileUtils.rm_r("tmp/test-pipes")
rescue Errno::ENOENT
end
FileUtils.mkdir_p("tmp/test-pipes/")
end
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
def rerun_failed_examples(failed_examples)
command = [
"bundle",
"exec",
"rspec",
"--format",
"documentation",
"--format",
"TurboTests::Flaky::FlakyDetectorFormatter",
*Flaky::Manager.potential_flaky_tests,
]
system(*command)
end
def start_multisite_subprocess(tests, **opts)
start_subprocess({}, %w[--tag type:multisite], tests, "multisite", **opts)
end
def start_regular_subprocess(tests, process_id, **opts)
start_subprocess(
{ "TEST_ENV_NUMBER" => process_id.to_s },
%w[--tag ~type:multisite],
tests,
process_id,
**opts,
)
end
def start_subprocess(env, extra_args, tests, process_id, record_runtime:)
if tests.empty?
@messages << { type: "exit", process_id: process_id }
else
tmp_filename = "tmp/test-pipes/subprocess-#{process_id}"
begin
File.mkfifo(tmp_filename)
rescue Errno::EEXIST
end
env["RSPEC_SILENCE_FILTER_ANNOUNCEMENTS"] = "1"
record_runtime_options =
if record_runtime
%w[--format ParallelTests::RSpec::RuntimeLogger --out tmp/turbo_rspec_runtime.log]
else
[]
end
command = [
"bundle",
"exec",
"rspec",
*extra_args,
"--order",
"random:#{@seed}",
"--format",
"TurboTests::JsonRowsFormatter",
"--out",
tmp_filename,
*record_runtime_options,
*tests,
]
env["DISCOURSE_RSPEC_PROFILE_EACH_EXAMPLE"] = "1" if @profile
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
command_string = [env.map { |k, v| "#{k}=#{v}" }.join(" "), command.join(" ")].join(" ")
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
if @verbose
STDOUT.puts "::group::[#{process_id}] Run RSpec" if ENV["GITHUB_ACTIONS"]
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
STDOUT.puts "Process #{process_id}: #{command_string}"
STDOUT.puts "::endgroup::" if ENV["GITHUB_ACTIONS"]
end
stdin, stdout, stderr, wait_thr = Open3.popen3(env, *command)
stdin.close
@threads << Thread.new do
File.open(tmp_filename) do |fd|
fd.each_line do |line|
message = JSON.parse(line)
message = message.symbolize_keys
message[:process_id] = process_id
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
message[:command_string] = command_string
@messages << message
end
end
@messages << { type: "exit", process_id: process_id }
end
@threads << start_copy_thread(stdout, STDOUT)
@threads << start_copy_thread(stderr, STDERR)
@threads << Thread.new { @messages << { type: "error" } if wait_thr.value.exitstatus != 0 }
end
end
def start_copy_thread(src, dst)
Thread.new do
while true
begin
msg = src.readpartial(4096)
rescue EOFError
src.close
break
else
dst.write(msg)
end
end
end
end
def handle_messages
exited = 0
begin
while true
message = @messages.pop
case message[:type]
when "example_passed"
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
example =
FakeExample.from_obj(
message[:example],
process_id: message[:process_id],
command_string: message[:command_string],
)
@reporter.example_passed(example)
when "example_pending"
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
example =
FakeExample.from_obj(
message[:example],
process_id: message[:process_id],
command_string: message[:command_string],
)
@reporter.example_pending(example)
when "example_failed"
DEV: Introduce automatic reruns to RSpec tests on Github actions (#24811) What motivated this change? Our builds on Github actions have been extremely flaky mostly due to system tests. This has led to a drop in confidence in our test suite where our developers tend to assume that a failed job is due to a flaky system test. As a result, we have had occurrences where changes that resulted in legitimate test failures are merged into the `main` branch because developers assumed it was a flaky test. What does this change do? This change seeks to reduce the flakiness of our builds on Github Actions by automatically re-running RSpec tests once when they fail. If a failed test passes subsequently in the re-run, we mark the test as flaky by logging it into a file on disk which is then uploaded as an artifact of the Github workflow run. We understand that automatically re-runs will lead to lower accuracy of our tests but we accept this as an acceptable trade-off since a fragile build has a much greater impact on our developers' time. Internally, the Discourse development team will be running a service to fetch the flaky tests which have been logged for internal monitoring. How is the change implemented? 1. A `--retry-and-log-flaky-tests` CLI flag is added to the `bin/turbo_rspec` CLI which will then initialize `TurboTests::Runner` with the `retry_and_log_flaky_tests` kwarg set to `true`. 2. When the `retry_and_log_flaky_tests` kwarg is set to `true` for `TurboTests::Runner`, we will register an additional formatter `Flaky::FailuresLoggerFormatter` to the `TurboTests::Reporter` in the `TurboTests::Runner#run` method. The `Flaky::FailuresLoggerFormatter` has a simple job of logging all failed examples to a file on disk when running all the tests. The details of the failed example which are logged can be found in `TurboTests::Flaky::FailedExample.to_h`. 3. Once all the tests have been run once, we check the result for any failed examples and if there are, we read the file on disk to fetch the `location_rerun_location` of the failed examples which is then used to run the tests in a new RSpec process. In the rerun, we configure a `TurboTests::Flaky::FlakyDetectorFormatter` with RSpec which removes all failed examples from the log file on disk since those examples are not flaky tests. Note that if there are too many failed examples on the first run, we will deem the failures to likely not be due to flaky tests and not re-run the test failures. As of writing, the threshold of failed examples is set to 10. If there are more than 10 failed examples, we will not re-run the failures.
2023-12-12 18:18:27 -05:00
example =
FakeExample.from_obj(
message[:example],
process_id: message[:process_id],
command_string: message[:command_string],
)
@reporter.example_failed(example)
@failure_count += 1
if fail_fast_met
@threads.each(&:kill)
break
end
when "message"
@reporter.message(message[:message])
when "seed"
when "close"
when "error"
@reporter.error_outside_of_examples
@error = true
when "exit"
exited += 1
if @reporter.formatters.any? { |f| f.is_a?(DocumentationFormatter) }
@reporter.message("[#{message[:process_id]}] DONE (#{exited}/#{@num_processes + 1})")
end
break if exited == @num_processes + 1
else
STDERR.puts("Unhandled message in main process: #{message}")
end
STDOUT.flush
end
rescue Interrupt
end
end
def fail_fast_met
!@fail_fast.nil? && @failure_count >= @fail_fast
end
end
end