invalid ELF header when using flask-bcrypt

AlanSwenson commented 5 years ago

Context

ran zappa update in which the only thing I changed was adding "cors": true to zappa_Settings.json and now im getting invalid ELF header because of Bcrypt? causing the deployment to fail

Expected Behavior

normal update

Actual Behavior

[1545000508959] Instancing.. [1545000512152] bcrypt is required to use Flask-Bcrypt [1545000512153] /var/task/bcrypt/_bcrypt.abi3.so: invalid ELF header: ImportError

Steps to Reproduce

import bcrypt into a flask project?

Your Environment

Zappa version used: 0.47.1
Operating System and Python version: OSX Mojave, Python 3.6

Things I've tried

removing the offending "cors": true undeploying and deploying adding bcrypt to my pipfile

AlanSwenson commented 5 years ago

Figured it out. there is no manylinux wheels for bcrypt 3.1.5 which apparently came out the same day I submitted this. Downgraded to 3.1.4 and now it works again.

cl1ent commented 5 years ago

another solution that allows using the current bcrypt version is compiling your own packages by adding "use_precompiled_packages": false to your zappa_settings.json

this compiles the packages on your machine (or ci environment) and does not use the old versions from Miserlou/lambda-packages

make sure to compile the stuff on a suitable python version and with glibc, musl libc didn't work for me...

suminb commented 5 years ago

It appears bcrypt 3.1.5 and above does not provide pre-compiled wheels for manylinux/Python3.6. I had the same issue, and had to revert bcrypt back to 3.1.4 :-/

See:

charlax commented 5 years ago

I had pinned bcrypt to 3.1.4 but this came up again... Not sure why.

prattcmp commented 5 years ago

Still having this problem, even with bcrypt 3.1.4

bendog commented 5 years ago

I've just updated a previously working python3.6 zappa deployment to python3.7 with the latest zappa and i'm now getting this error too... i'm assuming the version of bcrypt packaged has been updated and we will need to pin a different version

charlax commented 5 years ago

My solution was to use a dockerized env to deploy from... This way I don't need to pin packages anymore.

prattcmp commented 5 years ago

@charlax that seems more like a workaround than a fix, but I may have to do that. What docker image did you use?

charlax commented 5 years ago

You're right, but I'm not sure providing prebuilt up-to-date packages is a core feature of Zappa, so it might make sense anyway, depending on your needs.

prattcmp commented 5 years ago

I've spent hours trying to configure Docker to work seamlessly and efficiently with Zappa. The current implementation by Zappa is a hack at best, making it unreasonable to use. Really wanted Zappa to just work instead of causing more headaches than its worth 👎

bendog commented 5 years ago

Fails lambda runtime = python3.7

Python 3.7.3 with pipenv zappa==0.48.2

tried bcrypt==3.1.3 bcrypt==3.1.4 bcrypt==3.1.5 bcrypt==3.1.6

Works lambda runtime = python3.6 <-- python3.7 fails

Python 3.6.8 with pipenv zappa==0.48.2 bcrypt==3.1.4

prattcmp commented 5 years ago

Finally got an implementation that I'm satisfied with using Docker and PyCharm. It creates the virtual environment outside of the bind mount because I was having issues whenever the container was killed while the virtual environment was still running. This alleviates those headaches, and all the other headaches I've been running into quite frankly.

Dockerfile:

FROM lambci/lambda:build-python3.7

WORKDIR /var/task

# Fancy prompt to remind you are in zappashell
RUN echo 'export PS1="\[\e[36m\]zappa>\[\e[m\] "' >> /root/.bashrc

# Additional RUN commands here
# RUN yum clean all && \
#    yum -y install <stuff>

COPY entrypoint.sh /entrypoint.sh
RUN chmod +x /entrypoint.sh

SHELL ["/bin/bash", "-c"]
ENTRYPOINT source /entrypoint.sh

entrypoint.sh:

#!/usr/bin/env bash

# Enter the virtual environment
echo "CREATING VIRTUAL ENVIRONMENT... PLEASE WAIT..."
python -m venv /var/env
source /var/env/bin/activate

# Install required Python packages (via pip)
echo "INSTALLING PIP PACKAGES..."
pip install -r requirements.txt

# Add source to bash
echo 'source /var/env/bin/activate' >> /root/.bashrc

# Start a new shell
$SHELL

fazla043264 commented 4 years ago

still have the same error. anyone found any solution?

Finally got an implementation that I'm satisfied with using Docker and PyCharm. It creates the virtual environment outside of the bind mount because I was having issues whenever the container was killed while the virtual environment was still running. This alleviates those headaches, and all the other headaches I've been running into quite frankly.

Dockerfile:
FROM lambci/lambda:build-python3.7

WORKDIR /var/task

# Fancy prompt to remind you are in zappashell
RUN echo 'export PS1="\[\e[36m\]zappa>\[\e[m\] "' >> /root/.bashrc

# Additional RUN commands here
# RUN yum clean all && \
#    yum -y install <stuff>

COPY entrypoint.sh /entrypoint.sh
RUN chmod +x /entrypoint.sh

SHELL ["/bin/bash", "-c"]
ENTRYPOINT source /entrypoint.sh
entrypoint.sh:
#!/usr/bin/env bash

# Enter the virtual environment
echo "CREATING VIRTUAL ENVIRONMENT... PLEASE WAIT..."
python -m venv /var/env
source /var/env/bin/activate

# Install required Python packages (via pip)
echo "INSTALLING PIP PACKAGES..."
pip install -r requirements.txt

# Add source to bash
echo 'source /var/env/bin/activate' >> /root/.bashrc

# Start a new shell
$SHELL

I have tried this solution but when I run my docker image I get the following error: Could not open requirements file: [Errno 2] No such file or directory: 'requirements.txt'

although I do have the file in the directory.

Miserlou / Zappa