def get_flo(self):
s3_client = boto3.client('s3')
debug("in get_flo")
try:
result = s3_client.get_object(Bucket=self.bucket_name,
Key=self.key_name)
except Exception as e:
text = repr(e)[:256]
self.add_error("failed to process s3 object {}/{} '{}'".format(
self.bucket_name, self.key_name, text))
# issue #14 - the below decode majik is from AWS sample code.
new_key = urllib.unquote_plus(self.key_name.encode('utf8'))
self.add_message("First get failed ({}), trying to unquote"
" ({})".format(self.key_name, new_key))
result = s3_client.get_object(Bucket=self.bucket_name, Key=new_key)
self.add_message("get_object worked after unescaping")
debug("after s3_client.get_object() result={}".format(type(result)))
if result['ContentLength'] > MAX_EXE_SIZE:
msg = """Too big: {}/{} {}
({})""".format(self.bucket_name, self.key_name,
result['ContentLength'], repr(result))
print(msg)
raise SigVerifyTooBig(msg)
debug("before body read")
flo = BytesIO(result['Body'].read())
debug("after read() flo={}".format(type(flo)))
return flo
def lambda_handler(event, context):
rss = event['rss']
bucket_name = event['bucket']
logging.info("Processing url: %s" % rss)
logging.info("Using bucket: %s" % bucket_name)
# session = boto3.Config(region_name="us-west-2")
polly = boto3.client("polly")
s3 = boto3.resource('s3')
bucket = s3.Bucket(bucket_name)
logging.info("getting list of existing objects in the given bucket")
files = set(o.key for o in bucket.objects.all())
feed = feedparser.parse(rss)
title = feed['feed']['title']
fg = FeedGenerator()
fg.load_extension('podcast')
fg.title('Audio podcast based on: %s' % title)
fg.link(href=feed.feed.link, rel='alternate')
fg.subtitle(feed.feed.description)
ENTRY_URL = "http://s3-{region}.amazonaws.com/{bucket}/{filename}"
for entry in get_entries(feed):
filename = "%s.mp3" % entry['id']
fe = fg.add_entry()
fe.id(entry['id'])
fe.title(entry['title'])
fe.published(entry['published'])
entry_url = ENTRY_URL.format(
bucket=bucket_name,
filename=filename,
region=os.environ["AWS_REGION_BUCKET"]
)
fe.enclosure(entry_url, 0, 'audio/mpeg')
if filename in files:
logging.info('Article "%s" with id %s already exist, skipping.'
% (entry['title'], entry['id']))
continue
try:
logging.info("Next entry, size: %d" % len(entry['content']))
logging.debug("Content: %s" % entry['content'])
response = polly.synthesize_speech(
Text=entry['content'],
OutputFormat="mp3",
VoiceId="Joanna"
)
with closing(response["AudioStream"]) as stream:
bucket.put_object(Key=filename, Body=stream.read())
except BotoCoreError as error:
logging.error(error)
bucket.put_object(Key='podcast.xml', Body=fg.rss_str(pretty=True))
def send_sns(self, msg, e=None, reraise=False):
# use first line of incoming msg as subject, but AWS limit is under 100
# chars.
# ASSUME anything over is due to long s3 URL and use heuristic
subject = msg.split('\n')[0]
if len(subject) >= 100:
# split assuming URL, then retain result (index 0) and file name
# (index -1). File name should be sufficient to allow page
# recipient to decide urgency of further investigation.
pieces = subject.split('/')
subject = "{} ... {}".format(pieces[0], pieces[-1])
if len(subject) >= 100:
# don't try to be smarter, full text is still in 'msg'
subject = "Truncated subject, examine message"
# append bucket & key, short key first
msg += "\n{}\nkey={}\nbucket={}".format(
os.path.basename(self.key_name), self.key_name, self.bucket_name)
# hack to get traceback in email
if e:
import traceback
msg += traceback.format_exc()
client = boto3.client("sns")
# keep a global to prevent infinite recursion on arn error
global topic_arn
topic_arn = os.environ.get('SNSARN', "")
if self.verbose:
print("snsarn: {}".format(topic_arn))
if not topic_arn:
# bad config, we expected this in the environ
# set flag so we don't re-raise
topic_arn = "no-topic-arn"
raise KeyError("Missing 'SNSARN' from environment")
try:
# if the publish fails, we still want to continue, so we get the
# details into the cloud watch logs. Otherwise, this can
# (sometimes) terminate the lambda causing retries & DLQ
response = client.publish(Message=msg,
Subject=subject,
TopicArn=topic_arn)
debug("sns publish: '{}'".format(response))
except Exception as e:
self.add_message("sns publish failed\n"
" msg ({}): '{}'\n"
" subj ({}): '{}'\n"
"exception: '{}'"
"".format(len(msg), str(msg), len(subject),
str(subject), str(e)))
def send_sns(self, msg, e=None, reraise=False):
# hack to get traceback in email
if e:
import traceback
msg += traceback.format_exc()
client = boto3.client("sns")
# keep a global to prevent infinite recursion on arn error
global topic_arn
topic_arn = os.environ.get('SNSARN', "")
if self.verbose:
print("snsarn: {}".format(topic_arn))
if not topic_arn:
# bad config, we expected this in the environ
# set flag so we don't re-raise
topic_arn = "no-topic-arn"
raise KeyError("Missing 'SNSARN' from environment")
response = client.publish(Message=msg,
TopicArn=topic_arn) # noqa: W0612
def get_flo(self):
s3_client = boto3.client('s3')
debug("in get_flo")
try:
# Make sure the object is really available taken from
# https://blog.rackspace.com/the-devnull-s3-bucket-hacking-with-aws-lambda-and-python
# Don't use defaults, though -- that's 100 sec during testing!
start_waiting = time.time()
waiter = s3_client.get_waiter('object_exists')
waiter.wait(Bucket=self.bucket_name,
Key=self.key_name,
WaiterConfig={
'Delay': 3,
'MaxAttempts': 3
})
result = s3_client.get_object(Bucket=self.bucket_name,
Key=self.key_name)
except Exception as e:
debug("s3 exceptions type: {}".format(type(e)))
self.had_s3_error = True
text = repr(e)[:256]
self.add_error("failed to process s3 object {}/{} '{}'".format(
self.bucket_name, self.key_name, text))
raise
finally:
self.s3_wait_time = time.time() - start_waiting
debug("after s3_client.get_object() result={}".format(type(result)))
if result['ContentLength'] > MAX_EXE_SIZE:
msg = """Too big: {}/{} {}
({})""".format(self.bucket_name, self.key_name,
result['ContentLength'], repr(result))
print(msg)
raise SigVerifyTooBig(msg)
debug("before body read")
flo = BytesIO(result['Body'].read())
debug("after read() flo={}".format(type(flo)))
return flo
def get_user():
# This function doesn't have to be decorated, because the API call to IAM
# will be traced thanks to the monkey-patching.
iam = boto3.client('iam')
return dumps(iam.get_user(UserName="******"), default=json_serial)
