import json import logging import os import pytest from tests.integration_tests.clouds import ImageSpecification, IntegrationCloud from tests.integration_tests.conftest import get_validated_source LOG = logging.getLogger('integration_testing.test_upgrade') LOG_TEMPLATE = """\n\ === `systemd-analyze` before: {pre_systemd_analyze} === `systemd-analyze` after: {post_systemd_analyze} === `systemd-analyze blame` before (first 10 lines): {pre_systemd_blame} === `systemd-analyze blame` after (first 10 lines): {post_systemd_blame} === `cloud-init analyze show` before:') {pre_analyze_totals} === `cloud-init analyze show` after:') {post_analyze_totals} === `cloud-init analyze blame` before (first 10 lines): ') {pre_cloud_blame} === `cloud-init analyze blame` after (first 10 lines): ') {post_cloud_blame} """ UNSUPPORTED_INSTALL_METHOD_MSG = ( "Install method '{}' not supported for this test" ) USER_DATA = """\ #cloud-config hostname: SRU-worked """ def test_clean_boot_of_upgraded_package(session_cloud: IntegrationCloud): source = get_validated_source(session_cloud) if not source.installs_new_version(): pytest.skip(UNSUPPORTED_INSTALL_METHOD_MSG.format(source)) return # type checking doesn't understand that skip raises if (ImageSpecification.from_os_image().release == 'bionic' and session_cloud.settings.PLATFORM == 'lxd_vm'): # The issues that we see on Bionic VMs don't appear anywhere # else, including when calling KVM directly. It likely has to # do with the extra lxd-agent setup happening on bionic. # Given that we still have Bionic covered on all other platforms, # the risk of skipping bionic here seems low enough. pytest.skip("Upgrade test doesn't run on LXD VMs and bionic") return launch_kwargs = { 'image_id': session_cloud.released_image_id, } with session_cloud.launch( launch_kwargs=launch_kwargs, user_data=USER_DATA, ) as instance: # get pre values pre_hostname = instance.execute('hostname') pre_cloud_id = instance.execute('cloud-id') pre_result = instance.execute('cat /run/cloud-init/result.json') pre_network = instance.execute('cat /etc/netplan/50-cloud-init.yaml') pre_systemd_analyze = instance.execute('systemd-analyze') pre_systemd_blame = instance.execute('systemd-analyze blame') pre_cloud_analyze = instance.execute('cloud-init analyze show') pre_cloud_blame = instance.execute('cloud-init analyze blame') # Ensure no issues pre-upgrade assert not json.loads(pre_result)['v1']['errors'] log = instance.read_from_file('/var/log/cloud-init.log') assert 'Traceback' not in log assert 'WARN' not in log # Upgrade and reboot instance.install_new_cloud_init(source, take_snapshot=False) instance.execute('hostname something-else') instance.restart() assert instance.execute('cloud-init status --wait --long').ok # 'cloud-init init' helps us understand if our pickling upgrade paths # have broken across re-constitution of a cached datasource. Some # platforms invalidate their datasource cache on reboot, so we run # it here to ensure we get a dirty run. assert instance.execute('cloud-init init').ok # get post values post_hostname = instance.execute('hostname') post_cloud_id = instance.execute('cloud-id') post_result = instance.execute('cat /run/cloud-init/result.json') post_network = instance.execute('cat /etc/netplan/50-cloud-init.yaml') post_systemd_analyze = instance.execute('systemd-analyze') post_systemd_blame = instance.execute('systemd-analyze blame') post_cloud_analyze = instance.execute('cloud-init analyze show') post_cloud_blame = instance.execute('cloud-init analyze blame') # Ensure no issues post-upgrade assert not json.loads(pre_result)['v1']['errors'] log = instance.read_from_file('/var/log/cloud-init.log') assert 'Traceback' not in log assert 'WARN' not in log # Ensure important things stayed the same assert pre_hostname == post_hostname assert pre_cloud_id == post_cloud_id assert pre_result == post_result assert pre_network == post_network # Calculate and log all the boot numbers pre_analyze_totals = [ x for x in pre_cloud_analyze.splitlines() if x.startswith('Finished stage') or x.startswith('Total Time') ] post_analyze_totals = [ x for x in post_cloud_analyze.splitlines() if x.startswith('Finished stage') or x.startswith('Total Time') ] # pylint: disable=logging-format-interpolation LOG.info(LOG_TEMPLATE.format( pre_systemd_analyze=pre_systemd_analyze, post_systemd_analyze=post_systemd_analyze, pre_systemd_blame='\n'.join(pre_systemd_blame.splitlines()[:10]), post_systemd_blame='\n'.join(post_systemd_blame.splitlines()[:10]), pre_analyze_totals='\n'.join(pre_analyze_totals), post_analyze_totals='\n'.join(post_analyze_totals), pre_cloud_blame='\n'.join(pre_cloud_blame.splitlines()[:10]), post_cloud_blame='\n'.join(post_cloud_blame.splitlines()[:10]), )) @pytest.mark.ci @pytest.mark.ubuntu def test_subsequent_boot_of_upgraded_package(session_cloud: IntegrationCloud): source = get_validated_source(session_cloud) if not source.installs_new_version(): if os.environ.get('TRAVIS'): # If this isn't running on CI, we should know pytest.fail(UNSUPPORTED_INSTALL_METHOD_MSG.format(source)) else: pytest.skip(UNSUPPORTED_INSTALL_METHOD_MSG.format(source)) return # type checking doesn't understand that skip raises launch_kwargs = {'image_id': session_cloud.released_image_id} with session_cloud.launch(launch_kwargs=launch_kwargs) as instance: instance.install_new_cloud_init( source, take_snapshot=False, clean=False ) instance.restart() assert instance.execute('cloud-init status --wait --long').ok