CircleCI 2.0 private environment keys for public repos - firebase

I have a repo that is currently private. My Firebase deployment token is stored as an Env Var in the CircleCI GUI. The CircleCI 2.0 documentation clearly states
Do not add keys or secrets to a public CircleCI project
Also, from what I can find in the FAQ, a CircleCI project is made public if the associated GitHub repo is made public.
Now, I intend to open source the project on GitHub, will it make the Env Var in CircleCI visible to anyone since the CircleCI project will become public?
If the Env Var is publicly visible, what is an advised way to keep my deployment token hidden from others? Do I have to resort to a solution like GCP KMS?

A CircleCI Employee clarified this for me on their forums.
That warning is meant for the config[.yml]. You can use UI-based CircleCI environment variables safely.
If the repo/project is public, you'll just want to make sure that:
envars in forked builds are turned off in settings
you don't echo/print those variables to build output at all since that might be visible to the public
So it is possible to have a public repo (and thus public CircleCI project) with deployment keys safely configured in the CircleCI GUI.

It sounds like you're using CircleCI for your repo, but the open-source version of your project won't require deployment tools like Firebase. If that's right, then you should keep the Firebase deployment token wherever it's convenient and secure for you.
It also sounds like maybe you're just keeping the deployment token directly in your repo right now, since that's private. If you only have a handful of secrets in your repo, something like git-crypt might meet your needs. Once you hit a larger volume, you're probably going to want something centralized, and using Cloud KMS to encrypt secrets is one option.

Related

Firebase Functions Config: hide Secret in CLI itself

I need to set a secret in my functions config but I want to make sure, that this secret cannot be accessed by anyone, even if the person has access to the firebase project and thus the cli itself.
What I mean by that is, if I set the secret in my cli, I can then retrieve this secret firebase functions:config:get. There are people who might have access to the firebase cli of this project that should not be able to access this secret.
Is there a way to achieve this?
The functions configuration you're referring to is only saved on the local machine. It will not be checked into source control. Other developers who use the CLI in the same project, but on different machines, will not be able to see that configuration.
The configuration will be available in the deployed code, however. If each developer needs a fully isolated configuration at runtime, they should each have their own project to work with.

Firebase Remote Config - copy to another project

I have two projects for dev and prod. I want to be able to run a script to copy dev config to prod.
Firebase Remote Config has an API for programatically updating Remote Config. But as far as I can tell, you need to init admin with a project-specific service account. It seems like I would need two admin instances, but I'm not sure that's possible?
I'm wondering if someone has done this before and has an example script. Thanks!
See docs:
https://firebase.google.com/docs/remote-config/automate-rc
There is no Firebase Admin SDK for Flutter, so you'll have to implement this on a different platform that is supported. For a list of these platforms and instructions on setting it up, see the documentation on adding Firebase to a server.
For these platforms that the Firebase Admin SDK targets, you can create multiple instances of the FirebaseApp class, and initialize each of them with different credentials and project configuration. For examples of how to do this, see the documentation on initializing multiple apps.

Deploy via firebase-tools API without previous login

I'm currently building an open-source microservice that makes use of Firebase Database, Hosting & Functions. The plan is to pack everything in a single binary and distribute this. So users will have a hazzle-free, "bring your own Firebase project"-solution. They'll just have to download the binary and their Firebase secret key and can then create a user and deploy the service via CLI.
The problem is, that firebase-tools require a $FIREBASE_TOKEN when deploying via its API. So users would have to install firebase-tools in order to be able to generate that token and they would also have to store it (or re-generate it frequently).
Since I would like to provide a hazzle-free experience, I'd rather generate that token myself with the help of the secret key.
The question is: is this even possible? And if yes: how??
My workaround for this is to reflect the login- and logout-commands of the Firebase-CLI on my own binary's CLI. So the user won't have to install another tool.
To get the refresh_token I then read the data from the firebase-tools-configstore, that is located in the user folder. This feels a little dirty, like accessing a private API. But I couldn't come up with a better solution. And it works.

Separate dev and prod Firebase environment

I am considering using Firebase as MBaaS, however I couldn't find any reliable solution to the following problem:
I would like to set up two separate Firebase environments, one for development and one for production, but I don't want to do a manual copy of features (eg. remote configuration setup, notification rules, etc.) between the development and production environment.
Is there any tool or method I can rely on? Setting up remote configuration or notification rules from scratch can be a daunting task and too risky.
Any suggestions? Is there a better approach than having two separate environments?
Before you post another answer to the question which explains how to set up separate Firebase accounts: it is not the question, read it again. The question is: how to TRANSFER changes between separate dev and prod accounts or any better solution than manually copy between them.
If you are using firebase-tools there is a command firebase use which lets you set up which project you are using for firebase deploy
firebase use --add will bring up a list of your projects, select one and it will ask you for an alias. From there you can firebase use alias and firebase deploy will push to that project.
In my personal use, I have my-app and my-app-dev as projects in the Firebase console.
As everyone has pointed out - you need more than one project/database.
But to answer your question regarding the need to be able to copy settings/data etc from development to production. I had the exact same need. A few months in development and testing, I didn't want to manually copy the data.
My result was to backup the data to a storage bucket, and then restore it from there into the other database. It's a pretty crude way to do it - and I did a whole database backup/restore - but you might be able to look in that direction for a more controlled way. I haven't used it - it's very new - but this might be a solution: NPM Module firestore-export-import
Edit: Firestore backup/export/import info here Cloud Firestore Exporting and Importing Data
If you're using Firebase RTDB, and not Firestore - this documentation might help:
Firebase Automated Backups
You will need to set the permissions correctly to allow your production database access to the same storage bucket as your development.
Good luck.
I'm not currently using Firebase, but considering it like yourself. Looks like the way to go is to create a completely separate project on the console. There was a blogpost up recommending this on the old Firebase site, looks to be removed now though. https://web.archive.org/web/20160310115701/https://www.firebase.com/blog/2015-10-29-managing-development-environments.html
Also this discussion recommending same:
https://groups.google.com/forum/#!msg/firebase-talk/L7ajIJoHPcA/7dsNUTDlyRYJ
The way I did it:
I had 2 projects on firebase- one for DEV other for PROD
Locally my app also had 2 branches - one named DEV, the other named PROD
In my DEV branch I always have JSON file of DEV firebase project & likewise for PROD
This way I am not required to maintain my JSONs.
You will need to manage different build types
Follow this
First, create a new project at Firebase console, name id as YOURAPPNAME-DEV
Click "Add android app" button and create a new app. Name it com.yourapp.debug, for example. New google-services.json file will
be downloaded automatically
Under your project src directory create new directory with name "debug" and copy new google-services.json file here
In your module level build.gradle add this
debug {
applicationIdSuffix ".debug"
}
Now when you build a debug build google-services.json from "debug" folder will be used and when you will build in release mode google-services.json from module root directory will be considered.
I'm updating this answer based on information I just found.
Step 1
In firebase.google.com, create your multiple environments (i.e.; dev, staging, prod)
mysite-dev
mysite-staging
mysite-prod
Step 2
a. Move to the directly you want to be your default (i.e.; dev)
b. Run firebase deploy
c. Once deployed, run firebase use --add
d. An option will come up to select from the different projects you currently have.
Scroll to the project you want to add: mysite-staging, and select it.
e. You'll then be asked for an alias for that project. Enter staging.
Run items a-e again for prod and dev, so that each environment will have an alias
Know which environment you're in
Run firebase use
default (mysite-dev)
* dev (mysite-dev)
staging (mysite-staging)
prod (mysite-dev)
(one of the environments will have an asterisk to the left of it. That's the one you're currently in. It will also be highlighted in blue)
Switch between environments
Run firebase use staging or firebase use prod to move between them.
Once you're in the environment you want, run firebase deploy and your project will deploy there.
Here's a couple helpful links...
CLI Reference
Deploying to multiple environments
Hope this helps.
We chose to fire up instances of the new Firebase emulator on a local dev server for Test and UAT, leaving GCP out of the picture altogether. It's designed exactly for this use-case.
https://firebase.google.com/docs/emulator-suite
This blogpost describes a very simple approach with a debug and release build type.
In a nutshell:
Create a new App on Firebase for each build type using different application id suffix.
Configure your Android project with the latest JSON file.
Using applicationIdSuffix, change the Application Id to match the different Apps on Firebase depending on the build type.
=> see the blogpost for a detailed description.
If you want to use different build flavors, read this extensive blogpost from the official firebase blog. It contains a lot of valuable information.
Hope that helps!
To solve this for my situation I created three Firebase projects, each with the same Android project (i.e. same applicationId without using the applicationIdSuffix suggested by others). This resulted in three google-services.json files which I stored in my Continuous Integration (CI) server as custom environment variables. For each stage of the build (dev/staging/prod), I used the corresponding google-services.json file.
For the Firebase project associated with dev, in its Android project, I added the debug SHA certificate fingerprint. But for staging and prod I just have CI sign the APK.
Here is a stripped-down .gitlab-ci.yml that worked for this setup:
# This is a Gitlab Continuous Integration (CI) Pipeline definition
# Environment variables:
# - variables prefixed CI_ are Gitlab predefined environment variables (https://docs.gitlab.com/ee/ci/variables/predefined_variables.html)
# - variables prefixed GNDR_CI are Gitlab custom environment variables (https://docs.gitlab.com/ee/ci/variables/#creating-a-custom-environment-variable)
#
# We have three Firebase projects (dev, staging, prod) where the same package name is used across all of them but the
# debug signing certificate is only provided for the dev one (later if there are other developers, they can have their
# own Firebase project that's equivalent to the dev one). The staging and prod Firebase projects use real certificate
# signing so we don't need to enter a Debug signing certificate for them. We don't check the google-services.json into
# the repository. Instead it's provided at build time either on the developer's machine or by the Gitlab CI server
# which injects it via custom environment variables. That way the google-services.json can reside in the default
# location, the projects's app directory. The .gitlab-ci.yml is configured to copy the dev, staging, and prod equivalents
# of the google-servies.json file into that default location.
#
# References:
# https://firebase.googleblog.com/2016/08/organizing-your-firebase-enabled-android-app-builds.html
# https://stackoverflow.com/questions/57129588/how-to-setup-firebase-for-multi-stage-release
stages:
- stg_build_dev
- stg_build_staging
- stg_build_prod
jb_build_dev:
stage: stg_build_dev
image: jangrewe/gitlab-ci-android
cache:
key: ${CI_PROJECT_ID}-android
paths:
- .gradle/
script:
- cp ${GNDR_CI_GOOGLE_SERVICES_JSON_DEV_FILE} app/google-services.json
- ./gradlew :app:assembleDebug
artifacts:
paths:
- app/build/outputs/apk/
jb_build_staging:
stage: stg_build_staging
image: jangrewe/gitlab-ci-android
cache:
key: ${CI_PROJECT_ID}-android
paths:
- .gradle/
dependencies: []
script:
- cp ${GNDR_CI_GOOGLE_SERVICES_JSON_STAGING_FILE} app/google-services.json
- ./gradlew :app:assembleDebug
artifacts:
paths:
- app/build/outputs/apk/
jb_build_prod:
stage: stg_build_prod
image: jangrewe/gitlab-ci-android
cache:
key: ${CI_PROJECT_ID}-android
paths:
- .gradle/
dependencies: []
script:
- cp ${GNDR_CI_GOOGLE_SERVICES_JSON_PROD_FILE} app/google-services.json
# GNDR_CI_KEYSTORE_FILE_BASE64_ENCODED created on Mac via:
# base64 --input ~/Desktop/gendr.keystore --output ~/Desktop/keystore_base64_encoded.txt
# Then the contents of keystore_base64_encoded.txt were copied and pasted as a Gitlab custom environment variable
# For more info see http://android.jlelse.eu/android-gitlab-ci-cd-sign-deploy-3ad66a8f24bf
- cat ${GNDR_CI_KEYSTORE_FILE_BASE64_ENCODED} | base64 --decode > gendr.keystore
- ./gradlew :app:assembleRelease
-Pandroid.injected.signing.store.file=$(pwd)/gendr.keystore
-Pandroid.injected.signing.store.password=${GNDR_CI_KEYSTORE_PASSWORD}
-Pandroid.injected.signing.key.alias=${GNDR_CI_KEY_ALIAS}
-Pandroid.injected.signing.key.password=${GNDR_CI_KEY_PASSWORD}
artifacts:
paths:
- app/build/outputs/apk/
I'm happy with this solution because it doesn't rely on build.gradle tricks which I believe are too opaque and thus hard to maintain. For example, when I tried the approaches using applicationIdSuffix and different buildTypes I found that I couldn't get instrumented tests to run or even compile when I tried to switch build types using testBuildType. Android seemed to give special properties to the debug buildType which I couldn't inspect to understand.
Virtuously, CI scrips though are quite transparent and easy to maintain, in my experience. Indeed, the approach I've described worked: When I ran each of the APKs generated by CI on an emulator, the Firebase console's "Run your app to verify installation" step went from
Checking if the app has communicated with our servers. You may need to uninstall and reinstall your app.
to:
Congratulations, you've successfully added Firebase to your app!
for all three apps as I started them one by one in an emulator.
Firebase has a page on this which goes through how to set it up for dev and prod
https://firebase.google.com/docs/functions/config-env
Set environment configuration for your project To store environment
data, you can use the firebase functions:config:set command in the
Firebase CLI. Each key can be namespaced using periods to group
related configuration together. Keep in mind that only lowercase
characters are accepted in keys; uppercase characters are not allowed.
For instance, to store the Client ID and API key for "Some Service",
you might run:
firebase functions:config:set someservice.key="THE API KEY" someservice.id="THE CLIENT ID"
Retrieve current environment configuration To inspect what's currently
stored in environment config for your project, you can use firebase
functions:config:get. It will output JSON something like this:
{
"someservice": {
"key":"THE API KEY",
"id":"THE CLIENT ID"
}
}
Create the Tow project with Dev and production Environment on the firebase
Download the json file from thre
and setup the SDK as per : https://firebase.google.com/docs/android/setup Or for Crashlytics: https://firebase.google.com/docs/crashlytics/get-started?platform=android
First, place the respective google_services.json for each buildType in the following locations:
app/src/debug/google_services.json
app/src/test/google_services.json
app/google_services.json
Note: Root app/google_services.json This file should be there according to the build variants copy the json code in the root json file
Now, let’s whip up some gradle tasks in your: app’s build.gradle to automate moving the appropriate google_services.json to app/google_services.json
copy this in the app/Gradle file
task switchToDebug(type: Copy) {
description = 'Switches to DEBUG google-services.json'
from "src/debug"
include "google-services.json"
into "."
}
task switchToRelease(type: Copy) {
description = 'Switches to RELEASE google-services.json'
from "src/release"
include "google-services.json"
into "."
}
Great — but having to manually run these tasks before you build your app is cumbersome. We would want the appropriate copy task above run sometime before: assembleDebug or :assembleRelease is run. Let’s see what happens when :assembleRelease is run: copy this one in the /gradlew file
Zaks-MBP:my_awesome_application zak$ ./gradlew assembleRelease
Parallel execution is an incubating feature.
.... (other tasks)
:app:processReleaseGoogleServices
....
:app:assembleRelease
Notice the :app:processReleaseGoogleServices task. This task is responsible for processing the root google_services.json file. We want the correct google_services.json to be processed, so we must run our copy task immediately beforehand.
Add this to your build.gradle. Note the afterEvaluate enclosing.
copy this in the app/Gradle file
afterEvaluate {
processDebugGoogleServices.dependsOn switchToDebug
processReleaseGoogleServices.dependsOn switchToRelease
}
Now, anytime :app:processReleaseGoogleServices is called, our newly defined :app:switchToRelease will be called beforehand. Same logic for the debug buildType. You can run :app:assembleRelease and the release version google_services.json will be automatically copied to your app module’s root folder.
The way we are doing it is by creating different json key files for different environments. We have used service account feature as recommended by google and have one development file and another for production

Running a Shiny App from GitHub

I have an R Shiny app that contains some sensitive information that I would not like to be made public, and do not want to pay for any services in order to get password authentication.
My question is if I create a private repository on GitHub with the server.R, ui.R and all supporting data and files, will anyone be able to run it with the runGitHub command (below) or in any way access my data?
library(shiny)
runGitHub("<private repository name>", "<my user name>")
if I create a private repository on GitHub with the server.R, ui.R and all supporting data and files, will anyone be able to run it with the runGitHub command ... or in any way access my data?
If the repository is private only people who have been granted access should be able to access it. This is true through the GitHub website as well as via direct Git access, which is almost certainly what runGitHub() does.
do not want to pay for any services in order to get password authentication
In general, private repositories on GitHub aren't free. Currently the cheapest plan that includes private repos is the Micro plan at $7 per month.
There are other Git hosting providers that do provide free private repositories. BitBucket and GitLab both come to mind.
It may be possible for you to get free private hosting on GitHub, e.g. if you are a student.
I have an R Shiny app that contains some sensitive information that I would not like to be made public
Finally, depending on the nature of the "sensitive information" you are trying to protect, there may be better options. It is fairly common to provide things like API keys and passwords as environment variables (especially when using PaaS providers like Heroku), or to commit "template" files like config.template.ini.

Resources