Merge pull request #273 from flatcar-linux/scripts

ci-container/test: add AWS test script for CI automation
This commit is contained in:
Krzesimir Nowak 2022-05-10 18:35:12 +02:00
commit a954d4f164
4 changed files with 382 additions and 1 deletions

View File

@ -102,3 +102,20 @@ DIGITALOCEAN_PARALLEL="${PARALLEL_TESTS:-8}"
VMWARE_ESX_PARALLEL="${PARALLEL_TESTS:-4}"
# VMWARE_ESX_CREDS should come from sdk_container/.env and must be
# base64-encoded.
# -- AWS --
: ${AWS_amd64_INSTANCE_TYPE:="t3.small"}
# Space separated list of instance types. On those instances the
# cl.internet kola test will be run if this test is selected to run.
: ${AWS_amd64_MORE_INSTANCE_TYPES:="m4.2xlarge"}
: ${AWS_arm64_INSTANCE_TYPE:="a1.large"}
# Space separated list of instance types. On those instances the
# cl.internet kola test will be run if this test is selected to run.
: ${AWS_arm64_MORE_INSTANCE_TYPES:=""}
: ${AWS_IAM_PROFILE:="ciauto-test"}
: ${AWS_REGION:="us-east-1"}
: ${AWS_AMI_ID:=""}
AWS_PARALLEL="${PARALLEL_TESTS:-8}"
# AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY should come from
# sdk_container/.env

View File

@ -146,6 +146,9 @@ function test_run() {
local success=false
# A job on each worker prunes old mantle images (docker image prune)
echo "docker rm -f '${container_name}'" >> ./ci-cleanup.sh
# Vendor tests may need to know if it is a first run or a rerun
touch "${work_dir}/first_run"
for retry in $(seq "${retries}"); do
local tapfile="results-run-${retry}.tap"
local failfile="failed-run-${retry}.txt"
@ -162,8 +165,9 @@ function test_run() {
\"${arch}\" \
\"${vernum}\" \
\"${tapfile}\" \
$@"
$*"
set -e
rm -f "${work_dir}/first_run"
docker run --pull always --rm --name="${container_name}" --privileged --net host -v /dev:/dev \
-w /work -v "$PWD":/work "${mantle_ref}" \

View File

@ -0,0 +1,82 @@
#!/bin/bash
# Copyright (c) 2022 The Flatcar Maintainers.
# Use of this source code is governed by a BSD-style license that can be
# found in the LICENSE file.
set -euo pipefail
# Test execution script for the AWS vendor image.
# This script is supposed to run in the mantle container.
source ci-automation/vendor_test.sh
board="${CIA_ARCH}-usr"
escaped_vernum="${CIA_VERNUM//+/-}"
image_name="ci-${escaped_vernum}-${CIA_ARCH}"
aws_instance_type_var="AWS_${CIA_ARCH}_INSTANCE_TYPE"
aws_instance_type="${!aws_instance_type_var}"
more_aws_instance_types_var="AWS_${CIA_ARCH}_MORE_INSTANCE_TYPES"
set -o noglob # there shouldn't be any instance types with asterisks
# in it, but…
more_aws_instance_types=( ${!more_aws_instance_types_var} )
set +o noglob
vmdk='flatcar_production_ami_vmdk_image.vmdk'
tarball="${vmdk}.bz2"
if [[ "${AWS_AMI_ID}" == "" ]]; then
if [[ -f "${vmdk}" ]]; then
echo "++++ ${CIA_TESTSCRIPT}: using existing ${vmdk} for ${CIA_VERNUM} (${CIA_ARCH}) ++++"
else
echo "++++ ${CIA_TESTSCRIPT}: downloading ${tarball} for ${CIA_VERNUM} (${CIA_ARCH}) ++++"
copy_from_buildcache "images/${CIA_ARCH}/${CIA_VERNUM}/${tarball}" .
lbunzip2 "${tarball}"
fi
aws_bucket="flatcar-kola-ami-import-${AWS_REGION}"
aws_s3_path="s3://${aws_bucket}/${escaped_vernum}/${board}/"
trap 'ore -d aws delete --region="${AWS_REGION}" --board="${board}" --name="${image_name}" --ami-name="${image_name}" --file="${vmdk}" --bucket "${aws_s3_path}"' EXIT
ore aws initialize --region="${AWS_REGION}" --bucket "${aws_bucket}"
AWS_AMI_ID=$(ore aws upload --force --region="${AWS_REGION}" --board="${board}" --name="${image_name}" --ami-name="${image_name}" --ami-description="Flatcar Test ${image_name}" --file="${vmdk}" --bucket "${aws_s3_path}" | jq -r .HVM)
echo "++++ ${CIA_TESTSCRIPT}: created new AMI ${AWS_AMI_ID} (will be removed after testing) ++++"
fi
run_kola_tests() {
local instance_type="${1}"; shift
local instance_tapfile="${1}"; shift
timeout --signal=SIGQUIT 6h \
kola run \
--board="${board}" \
--basename="${image_name}" \
--channel="${CIA_CHANNEL}" \
--offering='basic' \
--parallel="${AWS_PARALLEL}" \
--platform=aws \
--aws-ami="${AWS_AMI_ID}" \
--aws-region="${AWS_REGION}" \
--aws-type="${instance_type}" \
--aws-iam-profile="${AWS_IAM_PROFILE}" \
--tapfile="${instance_tapfile}" \
--torcx-manifest="${CIA_TORCX_MANIFEST}" \
"${@}"
}
query_kola_tests() {
shift; # ignore the instance type
kola list --platform=aws --filter "${@}"
}
# these are set in ci-config.env
export AWS_ACCESS_KEY_ID
export AWS_SECRET_ACCESS_KEY
run_kola_tests_on_instances \
"${aws_instance_type}" \
"${CIA_TAPFILE}" \
"${CIA_FIRST_RUN}" \
"${more_aws_instance_types[@]}" \
'--' \
'cl.internet' \
'--' \
"${@}"

View File

@ -51,6 +51,9 @@
# Path to the Torcx manifest. Usually passed to kola through the
# --torcx-manifest parameter.
#
# CIA_FIRST_RUN:
# 1 if this is a first run, 0 if it is a rerun of failed tests.
#
#
# After this script is sourced, the parameters in ${@} specify test
# cases / test case patterns to run.
@ -90,6 +93,11 @@ fi
ciavts_tapfile="${PWD}/${ciavts_work_dir}/${ciavts_tapfile}"
ciavts_torcx_manifest="${PWD}/${ciavts_main_work_dir}/torcx_manifest.json"
ciavts_first_run=0
if [[ -f "${ciavts_main_work_dir}/first_run" ]]; then
ciavts_first_run=1
fi
echo "++++ Running ${ciavts_testscript} inside ${ciavts_work_dir} ++++"
cd "${ciavts_work_dir}"
@ -102,6 +110,276 @@ CIA_TESTSCRIPT="${ciavts_testscript}"
CIA_GIT_VERSION="${ciavts_git_version}"
CIA_BUILD_TYPE="${ciavts_type}"
CIA_TORCX_MANIFEST="${ciavts_torcx_manifest}"
CIA_FIRST_RUN="${ciavts_first_run}"
# Unset all variables with ciavts_ prefix now.
unset -v "${!ciavts_@}"
# Prefixes all test names in the tap file with a given prefix, so the
# test name like "cl.basic" will become "extra-test.[${prefix}].cl.basic".
#
# Typical use:
# prefix_tap_file "${instance_type}" "${tapfile}"
#
# Parameters:
# 1 - prefix
# 2 - tap file, modified in place
function prefix_tap_file() {
local prefix="${1}"; shift
local tap_file="${1}"; shift
# drop the dots from prefix
local actual_prefix="extra-test.[${prefix}]."
sed --in-place --expression 's/^\(\s*\(not\)\?\s*ok[^-]*\s*-\s*\)\(\S\)/\1'"${actual_prefix}"'\3/g' "${tap_file}"
}
# Filters the test names, so it puts only the real names of the
# prefixed tests into the chosen variable. For example for prefix
# "foo", it will ignore the test name like "cl.basic", but will print
# "cl.internet" for a test name like "extra-test.[foo].cl.internet".
# "*" is treated specially - it will be inserted into the chosen
# variable if it is passed.
#
# Typical use:
# filter_prefixed_tests tests_to_run "${instance_type}" "${@}"
# if [[ "${#tests_to_run[@]}" -gt 0 ]]; then …; fi
#
# Parameters:
# 1 - name of an array variable where the filtering results will be stored
# 2 - prefix
# @ - test names
function filter_prefixed_tests() {
local var_name="${1}"; shift
local prefix="${1}"; shift
# rest of the parameters are test names
local -n results="${var_name}"
local name
local stripped_name
for name; do
stripped_name="${name#extra-test.\[${prefix}\].}"
if [[ "${stripped_name}" != "${name}" ]]; then
results+=( "${stripped_name}" )
continue
elif [[ "${name}" = '*' ]]; then
results+=( '*' )
fi
done
}
# Filters out the extra tests from the passed test names. Ignored test
# names begin with "extra-test.". The results of the filtering are
# inserted into the chosen variable.
#
# Typical use:
# filter_out_prefixed_tests tests_to_run "${@}"
# if [[ "${#tests_to_run[@]}" -gt 0 ]]; then …; fi
#
# Parameters:
# 1 - name of an array variable where the filtering results will be stored
# @ - test names
function filter_out_prefixed_tests() {
local var_name="${1}"; shift
local -n results="${var_name}"
local name
for name; do
if [[ "${name#extra-test.}" = "${name}" ]]; then
results+=( "${name}" )
fi
done
}
# Merges into the first (main) tap file the contents of other tap
# files. It is very simple - the function assumes that all the tap
# files begin with a line like:
#
# 1..${number_of_tests}
#
# Other lines that are processed should begin like:
#
# (not)? ok - ${test_name}
#
# Any other lines are copied verbatim.
#
# The other tap files should already be preprocessed by
# prefix_tap_file to avoid duplicated test names.
#
# Typical use:
# merge_tap_files "${tap_file}" extra-validation-*.tap
# rm -f extra-validation-*.tap
#
# Parameters:
# 1 - main tap file
# @ - other tap files
function merge_tap_files() {
local main_tap_file="${1}"; shift
# rest of the parameters are other tap files
local main_test_count=0
if [[ -f "${main_tap_file}" ]]; then
main_test_count=$(head --lines=1 "${main_tap_file}" | grep --only-matching '[0-9]\+$')
fi
local other_test_count
local other_tap_file
local tmp_tap_file="${main_tap_file}.mtf.tmp"
for other_tap_file; do
if [[ ! -f "${other_tap_file}" ]]; then
continue
fi
other_test_count=$(head --lines=1 "${other_tap_file}" | grep --only-matching '[0-9]\+$' || echo 0 )
((main_test_count+=other_test_count))
done
echo "1..${main_test_count}" >"${tmp_tap_file}"
if [[ -f "${main_tap_file}" ]]; then
tail --lines=+2 "${main_tap_file}" >>"${tmp_tap_file}"
fi
for other_tap_file; do
if [[ ! -f "${other_tap_file}" ]]; then
continue
fi
tail --lines=+2 "${other_tap_file}" >>"${tmp_tap_file}"
done
mv --force "${tmp_tap_file}" "${main_tap_file}"
}
# Runs or reruns the tests on the main instance and other
# instances. Other instances usually run a subset of tests only.
#
# For this function to work, the caller needs to define two functions
# beforehand:
#
# run_kola_tests that takes the following parameters:
# 1 - instance type
# 2 - tap file
# @ - tests to run
#
# query_kola_tests that takes the following parameters:
# 1 - instance type
# @ - tests to run
# This function should print the names of the tests to run. Every line
# of the output should have one test name to run. Any other cruft in
# the line will be ignored.
#
# Typical use:
# function run_kola_tests() {
# local instance_type="${1}"; shift
# local tap_file="${1}"; shift
# kola run … "${@}"
# }
#
# function query_kola_tests() {
# local instance_type="${1}"; shift
# kola list … "${@}"
# }
#
# args=(
# "${main_instance}"
# "${CIA_TAPFILE}"
# "${CIA_FIRST_RUN}"
# "${other_instance_types[@]}"
# '--'
# 'cl.internet'
# '--'
# "${tests_to_run[@]}"
# )
# run_kola_tests_on_instances "${args[@]}"
#
# Parameters:
# 1 - main instance type - there all the tests are being run
# 2 - main tap file
# 3 - if this is first run (1 if it is, 0 if it is a rerun)
# @ - other instance types followed by double dash (--) followed by
# test names for other instances to filter from the tests to be
# run followed by double dash, followed by tests to be run or
# rerun
function run_kola_tests_on_instances() {
local main_instance_type="${1}"; shift
local main_tapfile="${1}"; shift
local is_first_run="${1}"; shift
local other_instance_types=()
local other_tests=()
local arg
while [[ "${#}" -gt 0 ]]; do
arg="${1}"; shift
if [[ "${arg}" = '--' ]]; then
break
fi
other_instance_types+=( "${arg}" )
done
while [[ "${#}" -gt 0 ]]; do
arg="${1}"; shift
if [[ "${arg}" = '--' ]]; then
break
fi
other_tests+=( "${arg}" )
done
# rest of the parameters are tests to be run or rerun
local instance_type
local queried_tests
local instance_tests=()
local tests_on_instances_running=0
local other_tests_for_fgrep
other_tests_for_fgrep="$(printf '%s\n' "${other_tests[@]}")"
for instance_type in "${other_instance_types[@]}"; do
# On first run we usually pass the canonical test names like
# cl.basic, cl.internet or *, so we decide which tests should
# be run on the other instances based on this list. On the
# other hand, the rerun will contain names of the failed tests
# only, and those are specific - if a test failed on the main
# instance, the name of the test will be like cl.basic; if a
# test failed on other instance, the name of the test will be
# like extra-test.[…].cl.basic. So in case of reruns, we want
# to filter the extra tests first then we decide which tests
# should be run.
if [[ "${is_first_run}" -eq 1 ]]; then
set -o noglob # noglob should not be necessary, as
# query_kola_tests shouldn't return a
# wildcard, but better to be safe than sorry
queried_tests="$(query_kola_tests "${instance_type}" "${@}")"
instance_tests=( $(grep --only-matching --fixed-strings "${other_tests_for_fgrep}" <<<"${queried_tests}" || :) )
set +o noglob
else
filter_prefixed_tests instance_tests "${instance_type}" "${@}"
fi
if [[ "${#instance_tests[@]}" -gt 0 ]]; then
tests_on_instances_running=1
(
local instance_tapfile="instance_${instance_type}_validate.tap"
set +e
set -x
local output
output=$(run_kola_tests "${instance_type}" "${instance_tapfile}" "${instance_tests[@]}" 2>&1)
set +x
set -e
local escaped_instance_type
escaped_instance_type="$(sed -e 's/[\/&]/\\&/g' <<<"${instance_type}")"
printf "=== START ${instance_type} ===\n%s\n=== END ${instance_type} ===\n" "$(sed -e "s/^/${escaped_instance_type}: /g" <<<"${output}")"
prefix_tap_file "${instance_type}" "${instance_tapfile}"
) &
fi
done
local -a main_tests
filter_out_prefixed_tests main_tests "${@}"
if [[ "${#main_tests[@]}" -gt 0 ]]; then
# run in a subshell, so the set -x and set +e do not pollute
# the outer environment
(
set +e
set -x
run_kola_tests "${main_instance_type}" "${main_tapfile}" "${main_tests[@]}"
true
)
fi
if [[ "${tests_on_instances_running}" -eq 1 ]]; then
wait
merge_tap_files "${main_tapfile}" 'instance_'*'_validate.tap'
rm -f 'instance_'*'_validate.tap'
fi
}