File: //lib/python2.7/site-packages/azurelinuxagent/common/utils/archive.py
# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the Apache License.
import errno
import os
import re
import shutil
import zipfile
from datetime import datetime
import azurelinuxagent.common.logger as logger
from azurelinuxagent.common.utils import fileutil
# pylint: disable=W0105
"""
archive.py
The module supports the archiving of guest agent state. Guest
agent state is flushed whenever there is a incarnation change.
The flush is archived periodically (once a day).
The process works as follows whenever a new incarnation arrives.
1. Flush - move all state files to a new directory under
.../history/timestamp/.
2. Archive - enumerate all directories under .../history/timestamp
and create a .zip file named timestamp.zip. Delete the archive
directory
3. Purge - glob the list .zip files, sort by timestamp in descending
order, keep the first 50 results, and delete the rest.
... is the directory where the agent's state resides, by default this
is /var/lib/waagent.
The timestamp is an ISO8601 formatted value.
"""
# pylint: enable=W0105
ARCHIVE_DIRECTORY_NAME = 'history'
_MAX_ARCHIVED_STATES = 50
_CACHE_PATTERNS = [
re.compile(r"^VmSettings.\d+\.json$"),
re.compile(r"^(.*)\.(\d+)\.(agentsManifest)$", re.IGNORECASE),
re.compile(r"^(.*)\.(\d+)\.(manifest\.xml)$", re.IGNORECASE),
re.compile(r"^(.*)\.(\d+)\.(xml)$", re.IGNORECASE),
re.compile(r"waagent_status\.(\d+)\.json$")
]
_GOAL_STATE_PATTERN = re.compile(r"^(.*)/GoalState\.(\d+)\.xml$", re.IGNORECASE)
# Old names didn't have incarnation, new ones do. Ensure the regex captures both cases.
# 2018-04-06T08:21:37.142697_incarnation_N
# 2018-04-06T08:21:37.142697_incarnation_N.zip
_ARCHIVE_PATTERNS_DIRECTORY = re.compile(r"^\d{4}\-\d{2}\-\d{2}T\d{2}:\d{2}:\d{2}\.\d+(_incarnation_(\d+))?$$")
_ARCHIVE_PATTERNS_ZIP = re.compile(r"^\d{4}\-\d{2}\-\d{2}T\d{2}:\d{2}:\d{2}\.\d+(_incarnation_(\d+))?\.zip$")
class StateFlusher(object):
def __init__(self, lib_dir):
self._source = lib_dir
directory = os.path.join(self._source, ARCHIVE_DIRECTORY_NAME)
if not os.path.exists(directory):
try:
fileutil.mkdir(directory)
except OSError as exception:
if exception.errno != errno.EEXIST:
logger.error("{0} : {1}", self._source, exception.strerror)
def flush(self):
files = self._get_files_to_archive()
if not files:
return
archive_name = self._get_archive_name(files)
if archive_name is None:
return
if self._mkdir(archive_name):
self._archive(files, archive_name)
else:
self._purge(files)
def history_dir(self, name):
return os.path.join(self._source, ARCHIVE_DIRECTORY_NAME, name)
@staticmethod
def _get_archive_name(files):
"""
Gets the most recently modified GoalState.*.xml and uses that timestamp and incarnation for the archive name.
In a normal workflow, we expect there to be only one GoalState.*.xml at a time, but if the previous one
wasn't purged for whatever reason, we take the most recently modified goal state file.
If there are no GoalState.*.xml files, we return None.
"""
latest_timestamp_ms = None
incarnation = None
for current_file in files:
match = _GOAL_STATE_PATTERN.match(current_file)
if not match:
continue
modification_time_ms = os.path.getmtime(current_file)
if latest_timestamp_ms is None or latest_timestamp_ms < modification_time_ms:
latest_timestamp_ms = modification_time_ms
incarnation = match.groups()[1]
if latest_timestamp_ms is not None and incarnation is not None:
return datetime.utcfromtimestamp(latest_timestamp_ms).isoformat() + "_incarnation_{0}".format(incarnation)
return None
def _get_files_to_archive(self):
files = []
for current_file in os.listdir(self._source):
full_path = os.path.join(self._source, current_file)
for pattern in _CACHE_PATTERNS:
match = pattern.match(current_file)
if match is not None:
files.append(full_path)
break
return files
def _archive(self, files, timestamp):
for current_file in files:
dst = os.path.join(self.history_dir(timestamp), os.path.basename(current_file))
shutil.move(current_file, dst)
def _purge(self, files):
for current_file in files:
os.remove(current_file)
def _mkdir(self, name):
directory = self.history_dir(name)
try:
fileutil.mkdir(directory, mode=0o700)
return True
except IOError as exception:
logger.error("{0} : {1}".format(directory, exception.strerror))
return False
# TODO: use @total_ordering once RHEL/CentOS and SLES 11 are EOL.
# @total_ordering first appeared in Python 2.7 and 3.2
# If there are more use cases for @total_ordering, I will
# consider re-implementing it.
class State(object):
def __init__(self, path, timestamp):
self._path = path
self._timestamp = timestamp
@property
def timestamp(self):
return self._timestamp
def delete(self):
pass
def archive(self):
pass
def __eq__(self, other):
return self._timestamp == other.timestamp
def __ne__(self, other):
return self._timestamp != other.timestamp
def __lt__(self, other):
return self._timestamp < other.timestamp
def __gt__(self, other):
return self._timestamp > other.timestamp
def __le__(self, other):
return self._timestamp <= other.timestamp
def __ge__(self, other):
return self._timestamp >= other.timestamp
class StateZip(State):
def delete(self):
os.remove(self._path)
class StateDirectory(State):
def delete(self):
shutil.rmtree(self._path)
def archive(self):
fn_tmp = "{0}.zip.tmp".format(self._path)
filename = "{0}.zip".format(self._path)
ziph = None
try:
# contextmanager for zipfile.ZipFile doesn't exist for py2.6, manually closing it
ziph = zipfile.ZipFile(fn_tmp, 'w')
for current_file in os.listdir(self._path):
full_path = os.path.join(self._path, current_file)
ziph.write(full_path, current_file, zipfile.ZIP_DEFLATED)
finally:
if ziph is not None:
ziph.close()
os.rename(fn_tmp, filename)
shutil.rmtree(self._path)
class StateArchiver(object):
def __init__(self, lib_dir):
self._source = os.path.join(lib_dir, ARCHIVE_DIRECTORY_NAME)
if not os.path.isdir(self._source):
try:
fileutil.mkdir(self._source, mode=0o700)
except IOError as exception:
if exception.errno != errno.EEXIST:
logger.error("{0} : {1}", self._source, exception.strerror)
def purge(self):
"""
Delete "old" archive directories and .zip archives. Old
is defined as any directories or files older than the X
newest ones.
"""
states = self._get_archive_states()
states.sort(reverse=True)
for state in states[_MAX_ARCHIVED_STATES:]:
state.delete()
def archive(self):
states = self._get_archive_states()
for state in states:
state.archive()
def _get_archive_states(self):
states = []
for current_file in os.listdir(self._source):
full_path = os.path.join(self._source, current_file)
match = _ARCHIVE_PATTERNS_DIRECTORY.match(current_file)
if match is not None:
states.append(StateDirectory(full_path, match.group(0)))
match = _ARCHIVE_PATTERNS_ZIP.match(current_file)
if match is not None:
states.append(StateZip(full_path, match.group(0)))
return states