debian-forge-composer/tools/libvirt_test.sh
Ondřej Budai bc1a7dc30c tests: fix journalctl killing once again
3a8c6c8a introduced a new logic for killing journalctl. Unfortunately, it
doesn't work properly. In ostree tests, multiple journalctls are spawned
but there can be only one trap active at a time. This caused all but the last
journalctls to hang indefinitely. Unfortunately, hanging background processes
is something that causes the GitLab CI to hang indefinitely as well.

This commit modifies the logic a bit: The trap is still set. However, there's
also an explicit kill of journalctl after the compose is finished. After the
process is successfully killed, the trap is removed.

Signed-off-by: Ondřej Budai <ondrej@budai.cz>
2021-06-10 07:40:36 +02:00

312 lines
9.6 KiB
Bash
Executable file

#!/bin/bash
set -euo pipefail
OSBUILD_COMPOSER_TEST_DATA=/usr/share/tests/osbuild-composer/
# Get OS data.
source /etc/os-release
ARCH=$(uname -m)
# Take the image type passed to the script or use qcow2 by default if nothing
# was passed.
IMAGE_TYPE=${1:-qcow2}
# Take the boot type passed to the script or use BIOS by default if nothing
# was passed.
BOOT_TYPE=${2:-bios}
# Select the file extension based on the image that we are building.
IMAGE_EXTENSION=$IMAGE_TYPE
if [[ $IMAGE_TYPE == 'openstack' ]]; then
IMAGE_EXTENSION=qcow2
fi
# Colorful output.
function greenprint {
echo -e "\033[1;32m${1}\033[0m"
}
# Start libvirtd and test it.
greenprint "🚀 Starting libvirt daemon"
sudo systemctl start libvirtd
sudo virsh list --all > /dev/null
# Set a customized dnsmasq configuration for libvirt so we always get the
# same address on bootup.
sudo tee /tmp/integration.xml > /dev/null << EOF
<network>
<name>integration</name>
<uuid>1c8fe98c-b53a-4ca4-bbdb-deb0f26b3579</uuid>
<forward mode='nat'>
<nat>
<port start='1024' end='65535'/>
</nat>
</forward>
<bridge name='integration' stp='on' delay='0'/>
<mac address='52:54:00:36:46:ef'/>
<ip address='192.168.100.1' netmask='255.255.255.0'>
<dhcp>
<range start='192.168.100.2' end='192.168.100.254'/>
<host mac='34:49:22:B0:83:30' name='vm' ip='192.168.100.50'/>
</dhcp>
</ip>
</network>
EOF
if ! sudo virsh net-info integration > /dev/null 2>&1; then
sudo virsh net-define /tmp/integration.xml
sudo virsh net-start integration
fi
# Allow anyone in the wheel group to talk to libvirt.
greenprint "🚪 Allowing users in wheel group to talk to libvirt"
WHEEL_GROUP=wheel
if [[ $ID == rhel ]]; then
WHEEL_GROUP=adm
fi
sudo tee /etc/polkit-1/rules.d/50-libvirt.rules > /dev/null << EOF
polkit.addRule(function(action, subject) {
if (action.id == "org.libvirt.unix.manage" &&
subject.isInGroup("${WHEEL_GROUP}")) {
return polkit.Result.YES;
}
});
EOF
# Set up variables.
TEST_UUID=$(uuidgen)
IMAGE_KEY=osbuild-composer-qemu-test-${TEST_UUID}
INSTANCE_ADDRESS=192.168.100.50
# Set up temporary files.
TEMPDIR=$(mktemp -d)
BLUEPRINT_FILE=${TEMPDIR}/blueprint.toml
COMPOSE_START=${TEMPDIR}/compose-start-${IMAGE_KEY}.json
COMPOSE_INFO=${TEMPDIR}/compose-info-${IMAGE_KEY}.json
# Check for the smoke test file on the AWS instance that we start.
smoke_test_check () {
# Ensure the ssh key has restricted permissions.
SSH_KEY=${OSBUILD_COMPOSER_TEST_DATA}keyring/id_rsa
SSH_OPTIONS=(-o StrictHostKeyChecking=no -o ConnectTimeout=5)
SMOKE_TEST=$(sudo ssh "${SSH_OPTIONS[@]}" -i "${SSH_KEY}" redhat@"${1}" 'cat /etc/smoke-test.txt')
if [[ $SMOKE_TEST == smoke-test ]]; then
echo 1
else
echo 0
fi
}
# Get the compose log.
get_compose_log () {
COMPOSE_ID=$1
LOG_FILE=${WORKSPACE}/osbuild-${ID}-${VERSION_ID}-${IMAGE_TYPE}.log
# Download the logs.
sudo composer-cli compose log "$COMPOSE_ID" | tee "$LOG_FILE" > /dev/null
}
# Get the compose metadata.
get_compose_metadata () {
COMPOSE_ID=$1
METADATA_FILE=${WORKSPACE}/osbuild-${ID}-${VERSION_ID}-${IMAGE_TYPE}.json
# Download the metadata.
sudo composer-cli compose metadata "$COMPOSE_ID" > /dev/null
# Find the tarball and extract it.
TARBALL=$(basename "$(find . -maxdepth 1 -type f -name "*-metadata.tar")")
tar -xf "$TARBALL"
rm -f "$TARBALL"
# Move the JSON file into place.
cat "${COMPOSE_ID}".json | jq -M '.' | tee "$METADATA_FILE" > /dev/null
}
# Write a basic blueprint for our image.
tee "$BLUEPRINT_FILE" > /dev/null << EOF
name = "bp"
description = "A base system"
version = "0.0.1"
EOF
# Prepare the blueprint for the compose.
greenprint "📋 Preparing blueprint"
sudo composer-cli blueprints push "$BLUEPRINT_FILE"
sudo composer-cli blueprints depsolve bp
# Get worker unit file so we can watch the journal.
WORKER_UNIT=$(sudo systemctl list-units | grep -o -E "osbuild.*worker.*\.service")
sudo journalctl -af -n 1 -u "${WORKER_UNIT}" &
WORKER_JOURNAL_PID=$!
# Stop watching the worker journal when exiting.
trap 'sudo pkill -P ${WORKER_JOURNAL_PID}' EXIT
# Start the compose
greenprint "🚀 Starting compose"
sudo composer-cli --json compose start bp "$IMAGE_TYPE" | tee "$COMPOSE_START"
COMPOSE_ID=$(jq -r '.build_id' "$COMPOSE_START")
# Wait for the compose to finish.
greenprint "⏱ Waiting for compose to finish: ${COMPOSE_ID}"
while true; do
sudo composer-cli --json compose info "${COMPOSE_ID}" | tee "$COMPOSE_INFO" > /dev/null
COMPOSE_STATUS=$(jq -r '.queue_status' "$COMPOSE_INFO")
# Is the compose finished?
if [[ $COMPOSE_STATUS != RUNNING ]] && [[ $COMPOSE_STATUS != WAITING ]]; then
break
fi
# Wait 30 seconds and try again.
sleep 5
done
# Capture the compose logs from osbuild.
greenprint "💬 Getting compose log and metadata"
get_compose_log "$COMPOSE_ID"
get_compose_metadata "$COMPOSE_ID"
# Kill the journal monitor immediately and remove the trap
sudo pkill -P ${WORKER_JOURNAL_PID}
trap - EXIT
# Did the compose finish with success?
if [[ $COMPOSE_STATUS != FINISHED ]]; then
echo "Something went wrong with the compose. 😢"
exit 1
fi
# Download the image.
greenprint "📥 Downloading the image"
# Current $PWD is inside /tmp, there may not be enough space for an image.
# Let's use a bigger temporary directory for this operation.
BIG_TEMP_DIR=/var/lib/osbuild-composer-tests
sudo rm -rf "${BIG_TEMP_DIR}" || true
sudo mkdir "${BIG_TEMP_DIR}"
pushd "${BIG_TEMP_DIR}"
sudo composer-cli compose image "${COMPOSE_ID}" > /dev/null
IMAGE_FILENAME=$(basename "$(find . -maxdepth 1 -type f -name "*.${IMAGE_EXTENSION}")")
LIBVIRT_IMAGE_PATH=/var/lib/libvirt/images/${IMAGE_KEY}.${IMAGE_EXTENSION}
sudo mv "$IMAGE_FILENAME" "$LIBVIRT_IMAGE_PATH"
popd
# Prepare cloud-init data.
CLOUD_INIT_DIR=$(mktemp -d)
cp "${OSBUILD_COMPOSER_TEST_DATA}"/cloud-init/{meta,user}-data "${CLOUD_INIT_DIR}"/
cp "${OSBUILD_COMPOSER_TEST_DATA}"/cloud-init/network-config "${CLOUD_INIT_DIR}"/
# Set up a cloud-init ISO.
greenprint "💿 Creating a cloud-init ISO"
CLOUD_INIT_PATH=/var/lib/libvirt/images/seed.iso
rm -f $CLOUD_INIT_PATH
pushd "$CLOUD_INIT_DIR"
sudo mkisofs -o $CLOUD_INIT_PATH -V cidata \
-r -J user-data meta-data network-config > /dev/null 2>&1
popd
# Ensure SELinux is happy with our new images.
greenprint "👿 Running restorecon on image directory"
sudo restorecon -Rv /var/lib/libvirt/images/
# Run virt-install to import the QCOW and boot it.
greenprint "🚀 Booting the image with libvirt"
if [[ $ARCH == 'ppc64le' ]]; then
# ppc64le has some machine quirks that must be worked around.
sudo virt-install \
--name "$IMAGE_KEY" \
--memory 2048 \
--vcpus 2 \
--disk path="${LIBVIRT_IMAGE_PATH}" \
--disk path=${CLOUD_INIT_PATH},device=cdrom \
--import \
--os-variant rhel8-unknown \
--noautoconsole \
--network network=integration,mac=34:49:22:B0:83:30 \
--qemu-commandline="-machine pseries,cap-cfpc=broken,cap-sbbc=broken,cap-ibs=broken,cap-ccf-assist=off,cap-large-decr=off"
elif [[ $ARCH == 's390x' ]]; then
# Our s390x machines are highly constrained on resources.
sudo virt-install \
--name "$IMAGE_KEY" \
--memory 512 \
--vcpus 1 \
--disk path="${LIBVIRT_IMAGE_PATH}" \
--disk path=${CLOUD_INIT_PATH},device=cdrom \
--import \
--os-variant rhel8-unknown \
--noautoconsole \
--network network=integration,mac=34:49:22:B0:83:30
else
# Both aarch64 and x86_64 support hybrid boot
if [[ $BOOT_TYPE == 'uefi' ]]; then
sudo virt-install \
--name "$IMAGE_KEY" \
--memory 1024 \
--vcpus 2 \
--disk path="${LIBVIRT_IMAGE_PATH}" \
--disk path=${CLOUD_INIT_PATH},device=cdrom \
--import \
--os-variant rhel8-unknown \
--noautoconsole \
--boot uefi,nvram_template=/usr/share/edk2/ovmf/OVMF_VARS.fd \
--network network=integration,mac=34:49:22:B0:83:30
else
sudo virt-install \
--name "$IMAGE_KEY" \
--memory 1024 \
--vcpus 2 \
--disk path="${LIBVIRT_IMAGE_PATH}" \
--disk path=${CLOUD_INIT_PATH},device=cdrom \
--import \
--os-variant rhel8-unknown \
--noautoconsole \
--network network=integration,mac=34:49:22:B0:83:30
fi
fi
# Set a number of maximum loops to check for our smoke test file via ssh.
case $ARCH in
s390x)
# s390x needs more time to boot its VM.
MAX_LOOPS=60
;;
*)
MAX_LOOPS=30
;;
esac
# Check for our smoke test file.
greenprint "🛃 Checking for smoke test file in VM"
# shellcheck disable=SC2034 # Unused variables left for readability
for LOOP_COUNTER in $(seq 0 ${MAX_LOOPS}); do
RESULTS="$(smoke_test_check $INSTANCE_ADDRESS)"
if [[ $RESULTS == 1 ]]; then
echo "Smoke test passed! 🥳"
break
fi
echo "Machine is not ready yet, retrying connection."
sleep 10
done
# Clean up our mess.
greenprint "🧼 Cleaning up"
sudo virsh destroy "${IMAGE_KEY}"
if [[ $ARCH == aarch64 || $BOOT_TYPE == 'uefi' ]]; then
sudo virsh undefine "${IMAGE_KEY}" --nvram
else
sudo virsh undefine "${IMAGE_KEY}"
fi
sudo rm -f "$LIBVIRT_IMAGE_PATH" $CLOUD_INIT_PATH
# Also delete the compose so we don't run out of disk space
sudo composer-cli compose delete "${COMPOSE_ID}" > /dev/null
# Use the return code of the smoke test to determine if we passed or failed.
if [[ $RESULTS == 1 ]]; then
greenprint "💚 Success"
else
greenprint "❌ Failed"
exit 1
fi
exit 0