I am deploying FastAPI app using PyTorch as requirements. The build fails most of the time and succeed when I remove PyTorch from requirements.txt.
Here is my configuration:
app.yaml
env: standard
runtime: python310
instance_class: F2
entrypoint: gunicorn -w 4 -k uvicorn.workers.UvicornWorker main:app
automatic_scaling:
target_cpu_utilization: 0.65
min_instances: 5
max_instances: 100
min_pending_latency: 30ms
max_pending_latency: automatic
max_concurrent_requests: 50
cloudbuild.yaml
# [START cloudbuild]
steps:
# Install dependencies
- name: 'python'
entrypoint: pip3
args: ['install', '-r', 'requirements.txt']
# Run custom commands
- name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
entrypoint: 'bash'
args: ['-c', 'gcloud config set app/cloud_build_timeout 1600 && gcloud app deploy']
timeout: '6000s'
# [END cloudbuild]
requirements.txt
torch
fastapi
uvicorn[standard]
gunicorn
sqlalchemy
google-api-python-client
google-auth-httplib2
google-auth-oauthlib
pydantic
transformers
cloud-sql-python-connector
pg8000
What should be changed to the configuration to make the deployment successful?
Related
I am trying to add Cloud Build on top of my App Engine Flask app. Everything works, but for some reason, I can't access the substitution variables I declared in the trigger.
Env vars are still being fetched from app.yaml. And they are parsed literally, not as variables. When I remove it from app.yaml Python throws a NoneType error.
[Trigger][1]: https://i.stack.imgur.com/Ii6Jv.png
[App.yaml][2]: https://i.stack.imgur.com/bg646.png
runtime: python310
instance_class: F4
automatic_scaling:
max_instances: 8
env_variables:
_CONFIG_TYPE: ${_CONFIG_TYPE}
[cloudbuild][3] https://i.stack.imgur.com/jo0PN.png
steps:
- name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
entrypoint: 'bash'
args: ['-c', 'gcloud config set app/cloud_build_timeout 1600 && gcloud app deploy']
timeout: '1600s'
substitutions:
_CONFIG_TYPE: ${_CONFIG_TYPE}
It won't work because gcloud app deploy command start a new Cloud Build, behind the scene, to build a container with your code and to deploy it. Your env var will change anything
The solution is to perform a bash replacement, with sed for instance.
app.yaml file
runtime: python310
instance_class: F4
automatic_scaling:
max_instances: 8
env_variables:
_CONFIG_TYPE: ##_CONFIG_TYPE##
Cloud Build Step with env var usage
steps:
- name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
entrypoint: 'bash'
env-vars:
- CONFIG_TYPE: ${_CONFIG_TYPE}
args:
- '-c'
- |
sed -i "s/##_CONFIG_TYPE##/$${CONFIG_TYPE}/g" app.yaml
gcloud config set app/cloud_build_timeout 1600
gcloud app deploy
timeout: '1600s'
substitutions:
_CONFIG_TYPE: ${_CONFIG_TYPE}
Cloud Build Step without env var usage
steps:
- name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
entrypoint: 'bash'
args:
- '-c'
- |
sed -i "s/##_CONFIG_TYPE##/${_CONFIG_TYPE}/g" app.yaml
gcloud config set app/cloud_build_timeout 1600
gcloud app deploy
timeout: '1600s'
substitutions:
_CONFIG_TYPE: ${_CONFIG_TYPE}
Below is my app.yaml
runtime: python39
entrypoint: gunicorn -b :$PORT main:app
runtime_config:
python_version: 3
env_variables:
SEC: %sec%
manual_scaling:
instances: 1
resources:
cpu: 1
memory_gb: 0.5
disk_size_gb: 10
This is my cloudbuild.yaml for app engine trying to pass a secrete vale to app.yaml as env veriable
steps:
- name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
entrypoint: 'bash'
args: ['-c', "export _VAL=$(echo $$SEC)
&& echo $$SEC;echo $_VAL && sed -i 's/%sec%/'$$SEC'/g' app.yaml
&& gcloud config set app/cloud_build_timeout 1600 && gcloud app deploy"
]
secretEnv: ["SEC"]
availableSecrets:
secretManager:
- versionName: projects/$PROJECT_ID/secrets/db_secret/versions/3
env: 'SEC'
timeout: '1600s'
I've done this in the past by just using ">>" to append the env vars to the bottom of my app.yaml. With this method, the env_variables section of your app.yaml needs to be last.
I don't use this method anymore though, the secrets show up in your cloudbuild log. I just import the secret manager package to grab my secrets inside the application these days.
cloudbuild.yaml
steps:
- name: "gcr.io/cloud-builders/gcloud"
secretEnv: ['SECRET_ONE','SECRET_TWO']
entrypoint: 'bash'
args:
- -c
- |
echo $'\n SECRET_ONE: '$$SECRET_ONE >> app.yaml
echo $'\n SECRET_TWO: '$$SECRET_TWO >> app.yaml
gcloud -q app deploy
availableSecrets:
secretManager:
- versionName: projects/012345678901/secrets/SECRET_ONE
env: 'SECRET_ONE'
- versionName: projects/012345678901/secrets/SECRET_TWO
env: 'SECRET_TWO'
app.yaml
runtime: go116
main: cmd
service: serviceone
env_variables:
PROJECT_ID: project-a0a00
PORT: 8080
You are in the right way, you have made a quote mistake because you try to run a inline command.
Prefer the | structure, more readable and therefore easy to debug and to write. Each line is executed sequentially, similar to the && in your inline command
steps:
- name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
entrypoint: 'bash'
args:
- '-c'
- |
export _VAL=$(echo $$SEC)
echo $$SEC
echo $_VAL
sed -i "s/%sec%/$$SEC/g" app.yaml
gcloud config set app/cloud_build_timeout 1600
gcloud app deploy
secretEnv: ["SEC"]
availableSecrets:
secretManager:
- versionName: projects/$PROJECT_ID/secrets/db_secret/versions/3
env: 'SEC'
timeout: '1600s'
EDIT 1:
If your password contain special characters that break the SED expression, you will have an error. You can use an alternative solution to replace the whole line, for example
steps:
- name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
entrypoint: 'bash'
args:
- '-c'
- |
export _VAL=$(echo $$SEC)
echo $$SEC
echo $_VAL
cat app.yaml | awk "{ if (NR == $(grep -n '%sec%' app.yaml | cut -d : -f 1)) print \" SEC: $$SEC\"; else print }" > app.yaml
gcloud config set app/cloud_build_timeout 1600
gcloud app deploy
secretEnv: ["SEC"]
availableSecrets:
secretManager:
- versionName: projects/$PROJECT_ID/secrets/db_secret/versions/3
env: 'SEC'
timeout: '1600s'
Have a try on it
I try to make a full process of integrated deployment with Gitlab and Google AppEngine.
But the deployment fail because it seems I miss rights. I really don't know which right is needed here.
I activated the App Engine API and the Cloud Build API.
The rights my service has:
my gitlab-ci.yml file:
Staging Deployment:
stage: Deploy
environment:
name: Staging
before_script:
- echo "deb http://packages.cloud.google.com/apt cloud-sdk-jessie main" | tee /etc/apt/sources.list.d/google-cloud-sdk.list
- curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | apt-key add -
- apt-get update
- apt-get -qq -y install google-cloud-sdk
script:
- echo $DEPLOY_KEY_FILE_PRODUCTION > /tmp/$CI_PIPELINE_ID.json
- gcloud auth activate-service-account --key-file /tmp/$CI_PIPELINE_ID.json
- gcloud config list
- gcloud config set project $PROJECT_ID_PRODUCTION
- gcloud --quiet --project $PROJECT_ID_PRODUCTION app deploy gae/test.yml
only:
refs:
- master
And finally my test.yml:
# [START django_app]
runtime: python37
service: testing
entrypoint: gunicorn -b :$PORT manta.wsgi --timeout 120
default_expiration: "5m"
env_variables:
DJANGO_SETTINGS_MODULE: manta.settings.dev
handlers:
# This configures Google App Engine to serve the files in the app's
# static directory.
- url: /static
static_dir: static/
# This handler routes all requests not caught above to the main app.
# It is required when static routes are defined, but can be omitted
# (along with the entire handlers section) when there are no static
# files defined.
- url: /.*
script: auto
# [END django_app]
And the error I have during deployment:
ERROR: (gcloud.app.deploy) Permissions error fetching application [apps/testing]. Please make sure you are using the correct project ID and that you have permission to view applications on the project.
I'm trying to dockerize a MERN stack but when the time comes when react has to start, the container exit with status 0.
This is the log error:
This is the structure of my project:
- project
- server
- api
api.yml
server.js
Dockerfile
- www
app.yml
Dockerfile
docker-compose.yml
The www folder contains the starter files the will be generate from npx create-react-app www.
The content of server/Dockerfile is:
FROM node:latest
RUN mkdir -p /usr/server
WORKDIR /usr/server
RUN npm install -g nodemon
EXPOSE 3000
CMD [ "npm", "start" ]
the content of www/Dockerfile is:
FROM node:latest
RUN mkdir -p /usr/www/src/app
WORKDIR /usr/www/src/app
EXPOSE 3000
CMD [ "npm", "start" ]
and, in the end the content of docker-compose.yml is:
version: '3.7'
services:
mongodb:
image: mongo
ports:
- 27017:27017
api:
build: ./server/
ports:
- "6200:6200"
volumes:
- ./server:/usr/server
depends_on:
- mongodb
www:
build: ./www/
ports:
- 3000:3000
volumes:
- ./www:/usr/www/src/app
depends_on:
- api
Now, in the title I mention Google Cloud, because I tried to distribute the different "server" and "www" parts in my production environment. The distribution of server works correctly butwww fails with this error:
The errors that are generated by Docker and Google Cloud seem very similar or am I wrong? Could it be a reaction problem or am I wrong in both cases?
I also leave the contents of the app.yaml andapi.yaml files.
app.yaml
runtime: nodejs
env: flex
# Only for developing
manual_scaling:
instances: 1
resources:
cpu: 1
memory_gb: 0.5
disk_size_gb: 10
handlers:
- url: /.*
static_files: build/index.html
upload: build/index.html
- url: /
static_dir: build
The content of the api.yaml file is the same as that of app.yaml but without the handlers section.
I am running command "gcloud app deploy" from my "vue" app folder and at the last getting vue-cli error, attached snap for it.
it seems like vue-cli-service isn't installed this is my cloudbuild.yaml if it helps
steps:
- name: node
entrypoint: npm
args: ['install']
# Necessary for vue.js projects
- name: node
entrypoint: npm
args: ['install', '-g', '#vue/cli']
- name: node
entrypoint: npm
args: ['test']
- name: node
entrypoint: npm
args: ['run', 'build']
- name: "gcr.io/cloud-builders/gcloud"
args: ["app", "deploy"]
app.yaml:
runtime: nodejs12
handlers:
# Serve all static files with urls ending with a file extension
- url: /(.*\..+)$
static_files: dist/\1
upload: dist/(.*\..+)$ # catch all handler to index.html
- url: /.*
static_files: dist/index.html
upload: dist/index.html
Note: if your package.json isn't in your repository root you should define set in npm commands as:
- name: node
entrypoint: npm
# --prefix and path:
args: ['install','--prefix','myapp']
and for deployment in gcloud if app.yaml isn't in root either:
- name: "gcr.io/cloud-builders/gcloud"
args: ["app", "deploy","myapp/app.yaml"]
Are you trying to deploy the app on GAE ( Google App Engine )?
If you are, the error might be caused because of GAE is not compatible with nodejs ( npm ), you need to deploy it in GAE flex if you want to use Vue of that way.
Another option could be, build the vue app and deploy it with another runtime, for example: python server; using the build as a static site.