def test_reboot_without_version_change(client: IntegrationInstance):
log = client.read_from_file("/var/log/cloud-init.log")
assert "Python version change detected" not in log
assert "Cache compatibility status is currently unknown." not in log
_assert_no_pickle_problems(log)
client.restart()
log = client.read_from_file("/var/log/cloud-init.log")
assert "Python version change detected" not in log
assert "Could not determine Python version used to write cache" not in log
_assert_no_pickle_problems(log)
# Now ensure that loading a bad pickle gives us problems
client.push_file(TEST_PICKLE, PICKLE_PATH)
client.restart()
log = client.read_from_file("/var/log/cloud-init.log")
# no cache found is an "expected" upgrade error, and
# "Failed" means we're unable to load the pickle
assert any(
[
"Failed loading pickled blob from {}".format(PICKLE_PATH) in log,
"no cache found" in log,
]
)
def test_dual_stack(client: IntegrationInstance):
# Drop IPv4 responses
assert client.execute("iptables -I INPUT -s 169.254.169.254 -j DROP").ok
_test_crawl(client, "http://[fd00:ec2::254]")
# Block IPv4 requests
assert client.execute("iptables -I OUTPUT -d 169.254.169.254 -j REJECT").ok
_test_crawl(client, "http://[fd00:ec2::254]")
# Re-enable IPv4
assert client.execute("iptables -D OUTPUT -d 169.254.169.254 -j REJECT").ok
assert client.execute("iptables -D INPUT -s 169.254.169.254 -j DROP").ok
# Drop IPv6 responses
assert client.execute("ip6tables -I INPUT -s fd00:ec2::254 -j DROP").ok
_test_crawl(client, "http://169.254.169.254")
# Block IPv6 requests
assert client.execute("ip6tables -I OUTPUT -d fd00:ec2::254 -j REJECT").ok
_test_crawl(client, "http://169.254.169.254")
# Force NoDHCPLeaseError (by removing dhclient) and assert ipv6 still works
# Destructive test goes last
# dhclient is at /sbin/dhclient on bionic but /usr/sbin/dhclient elseware
assert client.execute("rm $(which dhclient)").ok
client.restart()
log = client.read_from_file("/var/log/cloud-init.log")
assert "Crawl of metadata service using link-local ipv6 took" in log
def test_sudoers_includedir(client: IntegrationInstance):
"""Ensure we don't add additional #includedir to sudoers.
Newer versions of /etc/sudoers will use @includedir rather than
#includedir. Ensure we handle that properly and don't include an
additional #includedir when one isn't warranted.
https://github.com/canonical/cloud-init/pull/783
"""
if ImageSpecification.from_os_image().release in [
'xenial', 'bionic', 'focal'
]:
raise pytest.skip(
'Test requires version of sudo installed on groovy and later')
client.execute("sed -i 's/#include/@include/g' /etc/sudoers")
sudoers = client.read_from_file('/etc/sudoers')
if '@includedir /etc/sudoers.d' not in sudoers:
client.execute("echo '@includedir /etc/sudoers.d' >> /etc/sudoers")
client.instance.clean()
client.restart()
sudoers = client.read_from_file('/etc/sudoers')
assert '#includedir' not in sudoers
assert sudoers.count('includedir /etc/sudoers.d') == 1
def test_boot_event_disabled_by_default(client: IntegrationInstance):
log = client.read_from_file("/var/log/cloud-init.log")
if "network config is disabled" in log:
pytest.skip("network config disabled. Test doesn't apply")
assert "Applying network configuration" in log
assert "dummy0" not in client.execute("ls /sys/class/net")
_add_dummy_bridge_to_netplan(client)
client.execute("rm /var/log/cloud-init.log")
client.restart()
log2 = client.read_from_file("/var/log/cloud-init.log")
if "cache invalid in datasource" in log2:
# Invalid cache will get cleared, meaning we'll create a new
# "instance" and apply networking config, so events aren't
# really relevant here
pytest.skip("Test only valid for existing instances")
# We attempt to apply network config twice on every boot.
# Ensure neither time works.
assert 2 == len(
re.findall(r"Event Denied: scopes=\['network'\] EventType=boot[^-]",
log2))
assert 2 == log2.count(
"Event Denied: scopes=['network'] EventType=boot-legacy")
assert 2 == log2.count("No network config applied. Neither a new instance"
" nor datasource network update allowed")
assert "dummy0" in client.execute("ls /sys/class/net")
def _customize_envionment(client: IntegrationInstance):
# Assert our platform can detect LXD during sytemd generator timeframe.
ds_id_log = client.execute("cat /run/cloud-init/ds-identify.log").stdout
assert "check for 'LXD' returned found" in ds_id_log
# At some point Jammy will fail this test. We want to be informed
# when Jammy images no longer ship NoCloud template files (LP: #1958460).
assert "check for 'NoCloud' returned found" in ds_id_log
if client.settings.PLATFORM == "lxd_vm":
# ds-identify runs at systemd generator time before /dev/lxd/sock.
# Assert we can expected artifact which indicates LXD is viable.
result = client.execute("cat /sys/class/dmi/id/board_name")
if not result.ok:
raise AssertionError(
"Missing expected /sys/class/dmi/id/board_name")
if "LXD" != result.stdout:
raise AssertionError(f"DMI board_name is not LXD: {result.stdout}")
# Having multiple datasources prevents ds-identify from short-circuiting
# detection logic with a log like:
# single entry in datasource_list (LXD) use that.
# Also, NoCloud is detected during init-local timeframe.
# If there is a race on VMs where /dev/lxd/sock is not setup in init-local
# cloud-init will fallback to NoCloud and fail this test.
client.write_to_file(
"/etc/cloud/cloud.cfg.d/99-detect-lxd-first.cfg",
"datasource_list: [LXD, NoCloud]\n",
)
client.execute("cloud-init clean --logs")
client.restart()
def _check_iid_insensitive_across_kernel_upgrade(
instance: IntegrationInstance, ):
uuid = instance.read_from_file("/sys/class/dmi/id/product_uuid")
assert (uuid.isupper()
), "Expected uppercase UUID on Ubuntu FIPS image {}".format(uuid)
orig_kernel = instance.execute("uname -r").strip()
assert "azure-fips" in orig_kernel
result = instance.execute("apt-get update")
# Install a 5.4+ kernel which provides lowercase product_uuid
result = instance.execute("apt-get install linux-azure --assume-yes")
if not result.ok:
pytest.fail("Unable to install linux-azure kernel: {}".format(result))
# Remove ubuntu-azure-fips metapkg which mandates FIPS-flavour kernel
result = instance.execute("ua disable fips --assume-yes")
assert result.ok, "Unable to disable fips: {}".format(result)
instance.restart()
new_kernel = instance.execute("uname -r").strip()
assert orig_kernel != new_kernel
assert "azure-fips" not in new_kernel
assert "azure" in new_kernel
new_uuid = instance.read_from_file("/sys/class/dmi/id/product_uuid")
assert (
uuid.lower() == new_uuid
), "Expected UUID on linux-azure to be lowercase of FIPS: {}".format(uuid)
log = instance.read_from_file("/var/log/cloud-init.log")
RE_CONFIG_SSH_SEMAPHORE = r"Writing.*sem/config_ssh "
ssh_runs = len(re.findall(RE_CONFIG_SSH_SEMAPHORE, log))
assert 1 == ssh_runs, "config_ssh ran too many times {}".format(ssh_runs)
def _customize_envionment(client: IntegrationInstance):
client.write_to_file(
"/etc/cloud/cloud.cfg.d/99-detect-lxd.cfg",
"datasource_list: [LXD]\n",
)
client.execute("cloud-init clean --logs")
client.restart()
def _customize_environment(client: IntegrationInstance):
# Insert our "disable_network_activation" file here
client.write_to_file(
"/etc/cloud/cloud.cfg.d/99-disable-network-activation.cfg",
"disable_network_activation: true\n",
)
client.execute("cloud-init clean --logs")
client.restart()
def custom_client(client: IntegrationInstance,
tmpdir) -> Iterator[IntegrationInstance]:
client.write_to_file(f"/etc/cloud/cloud.cfg.d/{CUSTOM_CLOUD_DIR_FN}",
CUSTOM_CLOUD_DIR)
client.execute(f"rm -rf {DEFAULT_CLOUD_DIR}") # Remove previous cloud_dir
client.execute("cloud-init clean --logs")
client.restart()
yield client
def test_log_message_on_missing_version_file(client: IntegrationInstance):
# Start by pushing a pickle so we can see the log message
client.push_file(TEST_PICKLE, PICKLE_PATH)
client.execute("rm /var/lib/cloud/data/python-version")
client.restart()
log = client.read_from_file('/var/log/cloud-init.log')
assert ('Writing python-version file. '
'Cache compatibility status is currently unknown.') in log
def _customize_envionment(client: IntegrationInstance):
# Insert our "disable_network_config" file here
client.write_to_file(
"/etc/cloud/cloud.cfg.d/99-disable-network-config.cfg",
"network: {config: disabled}\n",
)
client.execute("cloud-init clean --logs")
client.restart()
def test_cache_purged_on_version_change(client: IntegrationInstance):
# Start by pushing the invalid pickle so we'll hit an error if the
# cache didn't actually get purged
client.push_file(TEST_PICKLE, PICKLE_PATH)
client.execute("echo '1.0' > /var/lib/cloud/data/python-version")
client.restart()
log = client.read_from_file("/var/log/cloud-init.log")
assert "Python version change detected. Purging cache" in log
_assert_no_pickle_problems(log)
def test_log_message_on_missing_version_file(client: IntegrationInstance):
client.push_file(TEST_PICKLE, PICKLE_PATH)
client.restart()
assert client.execute('cloud-init status --wait').ok
log = client.read_from_file('/var/log/cloud-init.log')
verify_ordered_items_in_text([
"Unable to unpickle datasource: 'MIMEMultipart' object has no "
"attribute 'policy'. Ignoring current cache.", 'no cache found',
'Searching for local data source',
'SUCCESS: found local data from DataSourceNoCloud'
], log)
def test_disk_setup_no_partprobe(create_disk, client: IntegrationInstance):
"""Ensure disk setup still works as expected without partprobe."""
# We can't do this part in a bootcmd because the path has already
# been found by the time we get to the bootcmd
client.execute('rm $(which partprobe)')
client.execute('cloud-init clean --logs')
client.restart()
log = client.read_from_file('/var/log/cloud-init.log')
_verify_first_disk_setup(client, log)
assert 'partprobe' not in log
def customize_environment(
client: IntegrationInstance,
tmpdir,
configure_secondary_nics: bool = False,
):
cfg = tmpdir.join("01_oracle_datasource.cfg")
with open(cfg, "w") as f:
f.write(
DS_CFG.format(configure_secondary_nics=configure_secondary_nics))
client.push_file(cfg, "/etc/cloud/cloud.cfg.d/01_oracle_datasource.cfg")
client.execute("cloud-init clean --logs")
client.restart()
def _test_network_config_applied_on_reboot(client: IntegrationInstance):
log = client.read_from_file('/var/log/cloud-init.log')
assert 'Applying network configuration' in log
assert 'dummy0' not in client.execute('ls /sys/class/net')
_add_dummy_bridge_to_netplan(client)
client.execute('rm /var/log/cloud-init.log')
client.restart()
log = client.read_from_file('/var/log/cloud-init.log')
assert 'Event Allowed: scope=network EventType=boot' in log
assert 'Applying network configuration' in log
assert 'dummy0' not in client.execute('ls /sys/class/net')
def test_disk_setup_when_mounted(
self, create_disk, client: IntegrationInstance
):
"""Test lp-1920939.
We insert an extra disk into our VM, format it to have two partitions,
modify our cloud config to mount devices before disk setup, and modify
our userdata to setup a single partition on the disk.
This allows cloud-init to attempt disk setup on a mounted partition.
When blockdev is in use, it will fail with
"blockdev: ioctl error on BLKRRPART: Device or resource busy" along
with a warning and a traceback. When partprobe is in use, everything
should work successfully.
"""
log = client.read_from_file("/var/log/cloud-init.log")
self._verify_first_disk_setup(client, log)
# Ensure NoCloud gets detected on reboot
client.execute("mkdir -p /var/lib/cloud/seed/nocloud-net/")
client.execute("touch /var/lib/cloud/seed/nocloud-net/meta-data")
client.write_to_file(
"/etc/cloud/cloud.cfg.d/99_nocloud.cfg",
"datasource_list: [ NoCloud ]\n",
)
# Update our userdata and cloud.cfg to mount then perform new disk
# setup
client.write_to_file(
"/var/lib/cloud/seed/nocloud-net/user-data",
UPDATED_PARTPROBE_USERDATA,
)
client.execute(
"sed -i 's/write-files/write-files\\n - mounts/' "
"/etc/cloud/cloud.cfg"
)
client.execute("cloud-init clean --logs")
client.restart()
# Assert new setup works as expected
verify_clean_log(log)
lsblk = json.loads(client.execute("lsblk --json"))
sdb = [x for x in lsblk["blockdevices"] if x["name"] == "sdb"][0]
assert len(sdb["children"]) == 1
assert sdb["children"][0]["name"] == "sdb1"
if "mountpoint" in sdb["children"][0]:
assert sdb["children"][0]["mountpoint"] == "/mnt3"
else:
assert sdb["children"][0]["mountpoints"] == ["/mnt3"]
def test_log_message_on_missing_version_file(client: IntegrationInstance):
# Start by pushing a pickle so we can see the log message
client.push_file(TEST_PICKLE, PICKLE_PATH)
client.execute("rm /var/lib/cloud/data/python-version")
client.execute("rm /var/log/cloud-init.log")
client.restart()
log = client.read_from_file("/var/log/cloud-init.log")
if "no cache found" not in log:
# We don't expect the python version file to exist if we have no
# pre-existing cache
assert (
"Writing python-version file. "
"Cache compatibility status is currently unknown." in log
)
def test_datasource_rbx_no_stacktrace(client: IntegrationInstance):
client.write_to_file(
"/etc/cloud/cloud.cfg.d/90_dpkg.cfg",
"datasource_list: [ RbxCloud, NoCloud, LXD ]\n",
)
client.write_to_file(
"/etc/cloud/ds-identify.cfg",
"policy: enabled\n",
)
client.execute("cloud-init clean --logs")
client.restart()
log = client.read_from_file("/var/log/cloud-init.log")
verify_clean_log(log)
assert "Failed to load metadata and userdata" not in log
assert ("Getting data from <class 'cloudinit.sources.DataSourceRbxCloud."
"DataSourceRbxCloud'> failed" not in log)
def test_reboot_without_version_change(client: IntegrationInstance):
log = client.read_from_file('/var/log/cloud-init.log')
assert 'Python version change detected' not in log
assert 'Cache compatibility status is currently unknown.' not in log
_assert_no_pickle_problems(log)
client.restart()
log = client.read_from_file('/var/log/cloud-init.log')
assert 'Python version change detected' not in log
assert 'Could not determine Python version used to write cache' not in log
_assert_no_pickle_problems(log)
# Now ensure that loading a bad pickle gives us problems
client.push_file(TEST_PICKLE, PICKLE_PATH)
client.restart()
log = client.read_from_file('/var/log/cloud-init.log')
assert 'Failed loading pickled blob from {}'.format(PICKLE_PATH) in log
def test_frequency_override(client: IntegrationInstance):
# Some pre-checks
assert ("running config-scripts-user with frequency once-per-instance"
in client.read_from_file("/var/log/cloud-init.log"))
assert client.read_from_file("/var/tmp/hi").strip().count("hi") == 1
# Change frequency of scripts-user to always
config = client.read_from_file("/etc/cloud/cloud.cfg")
new_config = config.replace("- scripts-user", "- [ scripts-user, always ]")
client.write_to_file("/etc/cloud/cloud.cfg", new_config)
client.restart()
# Ensure the script was run again
assert ("running config-scripts-user with frequency always"
in client.read_from_file("/var/log/cloud-init.log"))
assert client.read_from_file("/var/tmp/hi").strip().count("hi") == 2
def test_datasource_rbx_no_stacktrace(client: IntegrationInstance):
client.write_to_file(
'/etc/cloud/cloud.cfg.d/90_dpkg.cfg',
'datasource_list: [ RbxCloud, NoCloud ]\n',
)
client.write_to_file(
'/etc/cloud/ds-identify.cfg',
'policy: enabled\n',
)
client.execute('cloud-init clean --logs')
client.restart()
log = client.read_from_file('/var/log/cloud-init.log')
assert 'WARNING' not in log
assert 'Traceback' not in log
assert 'Failed to load metadata and userdata' not in log
assert ("Getting data from <class 'cloudinit.sources.DataSourceRbxCloud."
"DataSourceRbxCloud'> failed") not in log
def test_nocloud_seedfrom_vendordata(client: IntegrationInstance):
seed_dir = "/var/tmp/test_seed_dir"
result = client.execute(
"mkdir {seed_dir} && "
"touch {seed_dir}/user-data && "
"touch {seed_dir}/meta-data && "
"echo 'seedfrom: {seed_dir}/' > "
"/var/lib/cloud/seed/nocloud-net/meta-data".format(seed_dir=seed_dir))
assert result.return_code == 0
client.write_to_file(
"{}/vendor-data".format(seed_dir),
VENDOR_DATA,
)
client.execute("cloud-init clean --logs")
client.restart()
assert client.execute("cloud-init status").ok
assert "seeded_vendordata_test_file" in client.execute("ls /var/tmp")
def test_nocloud_seedfrom_vendordata(client: IntegrationInstance):
seed_dir = '/var/tmp/test_seed_dir'
result = client.execute(
"mkdir {seed_dir} && "
"touch {seed_dir}/user-data && "
"touch {seed_dir}/meta-data && "
"echo 'seedfrom: {seed_dir}/' > "
"/var/lib/cloud/seed/nocloud-net/meta-data".format(seed_dir=seed_dir)
)
assert result.return_code == 0
client.write_to_file(
'{}/vendor-data'.format(seed_dir),
VENDOR_DATA,
)
client.execute('cloud-init clean --logs')
client.restart(raise_on_cloudinit_failure=True)
assert 'seeded_vendordata_test_file' in client.execute('ls /var/tmp')
def test_disk_setup_when_mounted(
self, create_disk, client: IntegrationInstance
):
"""Test lp-1920939.
We insert an extra disk into our VM, format it to have two partitions,
modify our cloud config to mount devices before disk setup, and modify
our userdata to setup a single partition on the disk.
This allows cloud-init to attempt disk setup on a mounted partition.
When blockdev is in use, it will fail with
"blockdev: ioctl error on BLKRRPART: Device or resource busy" along
with a warning and a traceback. When partprobe is in use, everything
should work successfully.
"""
log = client.read_from_file('/var/log/cloud-init.log')
self._verify_first_disk_setup(client, log)
# Update our userdata and cloud.cfg to mount then perform new disk
# setup
client.write_to_file(
'/var/lib/cloud/seed/nocloud-net/user-data',
UPDATED_PARTPROBE_USERDATA,
)
client.execute(
"sed -i 's/write-files/write-files\\n - mounts/' "
"/etc/cloud/cloud.cfg"
)
client.execute('cloud-init clean --logs')
client.restart()
# Assert new setup works as expected
assert 'Traceback' not in log
assert 'WARN' not in log
lsblk = json.loads(client.execute('lsblk --json'))
sdb = [x for x in lsblk['blockdevices'] if x['name'] == 'sdb'][0]
assert len(sdb['children']) == 1
assert sdb['children'][0]['name'] == 'sdb1'
assert sdb['children'][0]['mountpoint'] == '/mnt3'
def _test_network_config_applied_on_reboot(client: IntegrationInstance):
log = client.read_from_file("/var/log/cloud-init.log")
if "network config is disabled" in log:
pytest.skip("network config disabled. Test doesn't apply")
assert "Applying network configuration" in log
assert "dummy0" not in client.execute("ls /sys/class/net")
_add_dummy_bridge_to_netplan(client)
client.execute('echo "" > /var/log/cloud-init.log')
client.restart()
log = client.read_from_file("/var/log/cloud-init.log")
if "cache invalid in datasource" in log:
# Invalid cache will get cleared, meaning we'll create a new
# "instance" and apply networking config, so events aren't
# really relevant here
pytest.skip("Test only valid for existing instances")
assert "Event Allowed: scope=network EventType=boot" in log
assert "Applying network configuration" in log
assert "dummy0" not in client.execute("ls /sys/class/net")
def test_boot_event_disabled_by_default(client: IntegrationInstance):
log = client.read_from_file('/var/log/cloud-init.log')
assert 'Applying network configuration' in log
assert 'dummy0' not in client.execute('ls /sys/class/net')
_add_dummy_bridge_to_netplan(client)
client.execute('rm /var/log/cloud-init.log')
client.restart()
log2 = client.read_from_file('/var/log/cloud-init.log')
# We attempt to apply network config twice on every boot.
# Ensure neither time works.
assert 2 == len(
re.findall(r"Event Denied: scopes=\['network'\] EventType=boot[^-]",
log2))
assert 2 == log2.count(
"Event Denied: scopes=['network'] EventType=boot-legacy")
assert 2 == log2.count("No network config applied. Neither a new instance"
" nor datasource network update allowed")
assert 'dummy0' in client.execute('ls /sys/class/net')
def _customize_envionment(client: IntegrationInstance):
# Assert our platform can detect LXD during systemd generator timeframe.
ds_id_log = client.execute("cat /run/cloud-init/ds-identify.log").stdout
assert "check for 'LXD' returned found" in ds_id_log
if client.settings.PLATFORM == "lxd_vm":
# ds-identify runs at systemd generator time before /dev/lxd/sock.
# Assert we can expected artifact which indicates LXD is viable.
result = client.execute("cat /sys/class/dmi/id/board_name")
if not result.ok:
raise AssertionError(
"Missing expected /sys/class/dmi/id/board_name")
if "LXD" != result.stdout:
raise AssertionError(f"DMI board_name is not LXD: {result.stdout}")
# Having multiple datasources prevents ds-identify from short-circuiting
# detection logic with a log like:
# single entry in datasource_list (LXD) use that.
# Also, NoCloud is detected during init-local timeframe.
# If there is a race on VMs where /dev/lxd/sock is not setup in init-local
# cloud-init will fallback to NoCloud and fail this test.
client.write_to_file(
"/etc/cloud/cloud.cfg.d/99-detect-lxd-first.cfg",
"datasource_list: [LXD, NoCloud]\n",
)
# This is also to ensure that NoCloud can be detected
if ImageSpecification.from_os_image().release == "jammy":
# Add nocloud-net seed files because Jammy no longer delivers NoCloud
# (LP: #1958460).
client.execute("mkdir -p /var/lib/cloud/seed/nocloud-net")
client.write_to_file("/var/lib/cloud/seed/nocloud-net/meta-data", "")
client.write_to_file("/var/lib/cloud/seed/nocloud-net/user-data",
"#cloud-config\n{}")
client.execute("cloud-init clean --logs")
client.restart()
