OSError: [Errno 28] No space left on this device on slim_handler=true

[thogerb]

I getting "No space left on this device" error with a vanilla Django project + TensorFlow. I have looked through the other issues on Zappa and TensorFlow, but with little luck.

Main question/confusion: Is the TensorFlow too big in itself for Zappa/Lambda or am I doing something wrong?

Hope you can help

Context

I am trying to use TensorFlow with Zappa, deploying a vanilla Django project and it's throwing the "No space left on this device". Reading through related issues, I don't understand why I am not able to get TensorFlow pushed to S3/Lambda when it is ONLY (no model):

• Zappa (and it's dependencies)
• Django
• TensorFlow (and it's depencies)

I have tried to go through all the issues here related to TensorFlow and it seems that people are able to get it up there in the cloud, but I have been struggling with it for quite some hours now.

Expected Behavior

Zappa should deploy without error.

Actual Behavior

Traceback (most recent call last):
File "/var/task/handler.py", line 602, in lambda_handler
return LambdaHandler.lambda_handler(event, context)
File "/var/task/handler.py", line 245, in lambda_handler
handler = cls()
File "/var/task/handler.py", line 102, in init
self.load_remote_project_archive(project_archive_path)
File "/var/task/handler.py", line 174, in load_remote_project_archive
t.extractall(project_folder)
File "/var/lang/lib/python3.6/tarfile.py", line 2010, in extractall
numeric_owner=numeric_owner)
File "/var/lang/lib/python3.6/tarfile.py", line 2052, in extract
numeric_owner=numeric_owner)
File "/var/lang/lib/python3.6/tarfile.py", line 2122, in _extract_member
self.makefile(tarinfo, targetpath)
File "/var/lang/lib/python3.6/tarfile.py", line 2171, in makefile
copyfileobj(source, target, tarinfo.size, ReadError, bufsize)
File "/var/lang/lib/python3.6/tarfile.py", line 252, in copyfileobj
dst.write(buf)
OSError: [Errno 28] No space left on device

Possible Fix

Steps to Reproduce

What I am doing:

1. Creating a new venv and only "pip install zappa django tensorflow"
2. Creating a new Django project
3. zappa init'ing (all defaults), adding slim_handler=true and zappa deploy dev'ing

Your Environment

• Zappa version used: 0.48.2
• Operating System and Python version: MacOs, Python 3.6

pip freeze

absl-py==0.7.1
argcomplete==1.9.3
astor==0.7.1
boto3==1.9.147
botocore==1.12.147
certifi==2019.3.9
cfn-flip==1.2.0
chardet==3.0.4
Click==7.0
Django==2.2.1
docutils==0.14
durationpy==0.5
future==0.16.0
gast==0.2.2
grpcio==1.20.1
h5py==2.9.0
hjson==3.0.1
idna==2.8
jmespath==0.9.3
kappa==0.6.0
Keras-Applications==1.0.7
Keras-Preprocessing==1.0.9
lambda-packages==0.20.0
Markdown==3.1
mock==3.0.5
numpy==1.16.3
placebo==0.9.0
protobuf==3.7.1
python-dateutil==2.6.1
python-slugify==1.2.4
pytz==2019.1
PyYAML==5.1
requests==2.21.0
s3transfer==0.2.0
six==1.12.0
sqlparse==0.3.0
tensorboard==1.13.1
tensorflow==1.13.1
tensorflow-estimator==1.13.0
termcolor==1.1.0
toml==0.10.0
tqdm==4.19.1
troposphere==2.4.6
Unidecode==1.0.23
urllib3==1.24.3
Werkzeug==0.15.2
wsgi-request-logger==0.4.6
zappa==0.48.2

pip package size of tensorflow

330M lib/python3.6/site-packages/tensorflow

[CodesInTheShell]

Are you able to solve this? I did "memory_size": 3008, but with no luck for a hello word flask app with autosklearn (sklearn, numpy, pandas, etc.)

[moviedatascience]

Were you ever able to solve this? I'm running into the same problem.

[thogerb]

Unfortunately not. I found some guidance online on how you might be able to strip down the Tensorflow package, for example getting rid of TensorBoard -- but never succeeded with it. Lambda sizes now don't seem to be a fit for the large bundle sized of numpy, sklearn, pandas, tensorflow, keras, and similar. Numpy alone is possible, but that also stretches capacity.

[moviedatascience]

Really unfortunate considering the gzip addition was supposedly created for tensorflow