search

broadinstitute / single_cell_portal_core

Rails/Docker application for the Broad Institute's single cell RNA-seq data portal
https://singlecell.broadinstitute.org
BSD 3-Clause "New" or "Revised" License
62 stars 26 forks source link
human-cell-atlas rna-seq-analysis single-cell-genomics

readme

SINGLE CELL PORTAL README

codecov Build Status

SETUP

This application is built and deployed using Docker, specifically native Docker for Mac OSX. Please refer to their online documentation for instructions on installing Docker.

PULLING/BUILDING THE DOCKER IMAGE

To pull the Docker image that Single Cell Portal is built off of, run the following command:

docker pull gcr.io/broad-singlecellportal-staging/single-cell-portal:development

This is not strictly necessary as a pre-flight step, but will make calling bin/boot_docker faster (see Running the Container) for more information.

To build the Docker image locally, run the following command from the project root directory:

docker build -t gcr.io/broad-singlecellportal-staging/single-cell-portal:development .

This will start the automated process of building the Docker image for running the portal. The image is based on the Passenger-docker baseimage and comes with Ruby, Nginx, and Passenger by default, but built on a different base image and with additional packages added to the Broad Institute SCP Rails baseimage which pulls from the original baseimage. The extended image contains other required libraries, and self-signed SSL certificates & CA Authority for doing local development in https.

If this is your first time building the image, it may take several minutes to download and install everything.

BEFORE RUNNING THE CONTAINER

This project uses MongoDB as the primary datastore, and connects to a remote MongoDB instance in all environments. In order for the portal to operate correctly, you will need to provision a MongoDB instance in Google Cloud Platform. This can be done using Google Click to Deploy, or via providers such as MongoDB Atlas. You may also provision a database manually using a VM in Google Compute Environment if you so desire.

The only requirement is that you add a user account called "single_cell" to your Single Cell Portal database in whatever instances you create.

EXTERNAL DEPENDENCIES AND INTEGRATIONS

The Single Cell Portal has several external dependencies & integrations that are either required for operation, or enable extra security/reporting features when certain variables are set. If you are deploying your own instance of the Single Cell Portal, you will need to create resources and accounts in order for the portal to function. They are as follows:

Google Cloud Platform (REQUIRED)

The Single Cell Portal is currently deployed in Google Cloud Platform (specifically, on a Google Compute Engine virtual machine). This is a requirement for all functionality. The portal utilizes GCP service accounts for making authenticated calls to various APIs as a part of normal usage. For more information on how to configure your GCP project, please see Local Development or Deploying a Private Instance below.

Terra (REQUIRED)

The Single Cell Portal uses Terra, which is the Broad Institute's cloud-based analysis platform. The portal uses Terra both as a data and permission store, as well as an analysis platform. For more information on the integration itself, see Terra Integration below, or for setting up your own Terra project, see Local Development or Deploying a Private Instance below.

Sentry

The Single Cell Portal is set up to integrate with Sentry, an external error reporting service. This is not required for the portal to function (this feature is opt-in and only functions when certain parameters are set at runtime). Developers deploying their own instance will need to register for an account with Sentry, and then set the SENTRY_DSN environment variable when deploying your instance (see Running the Container and DOCKER RUN COMMAND ENVIRONMENT VARIABLES for more detail).

Note - while the Sentry DSN is stored with secrets and may appear as one, it is not a secret and Sentry's official stance is that the Sentry DSN does not need to be kept private.

Google Analytics

The Single Cell Portal is configured to report web traffic to Google Analytics. You will first need to set up an account on Google Analytics, and then to enable simply set the GA_TRACKING_ID environment variable when deploying (see Running the Container and DOCKER RUN COMMAND ENVIRONMENT VARIABLES for more detail).

LOCAL DEVELOPMENT OR DEPLOYING A PRIVATE INSTANCE

If you are not part of the Single Cell Portal development team and are trying to use the portal locally or deploying a private instance of the Single Cell Portal (i.e. not any Broad Institute instance), there are a few extra steps that need to be taken before the portal is configured and ready to use:

RUNNING THE CONTAINER

Note: If you are developing locally and would like to leverage the hybrid Docker/local setup, see the section below on HYBRID DOCKER LOCAL DEVELOPMENT.

Once the image has successfully built, all registration/configuration steps have been completed, use the following command to start the container:

bin/boot_docker -u (sendgrid username) -P (sendgrid password) -k (service account key path) -K (read-only service account key path) -o (oauth client id) -S (oauth client secret) -N (portal namespace) -m (mongodb hostname) -M (mongodb internal ip) -p (mongodb password)

This sets up several environment variables in your shell and then runs the following command:

docker run --rm -it --name $CONTAINER_NAME -p 80:80 -p 443:443 -p 587:587 -h localhost -v $PROJECT_DIR:/home/app/webapp:rw -e PASSENGER_APP_ENV=$PASSENGER_APP_ENV -e MONGO_LOCALHOST=$MONGO_LOCALHOST -e MONGO_INTERNAL_IP=$MONGO_INTERNAL_IP -e SENDGRID_USERNAME=$SENDGRID_USERNAME -e SENDGRID_PASSWORD=$SENDGRID_PASSWORD -e SECRET_KEY_BASE=$SECRET_KEY_BASE -e PORTAL_NAMESPACE=$PORTAL_NAMESPACE -e SERVICE_ACCOUNT_KEY=$SERVICE_ACCOUNT_KEY -e OAUTH_CLIENT_ID=$OAUTH_CLIENT_ID -e OAUTH_CLIENT_SECRET=$OAUTH_CLIENT_SECRET -e SENTRY_DSN=$SENTRY_DSN -e GA_TRACKING_ID=$GA_TRACKING_ID gcr.io/broad-singlecellportal-staging/single-cell-portal:development

The container will then start running, and will execute its local startup scripts that will configure the application automatically.

You can also run the bin/boot_docker script in help mode by passing -H to print the help text which will show you how to pass specific values to the above env variables. Note: running the shortcut script with an environment of 'production' will cause the container to spawn headlessly by passing the -d flag, rather than --rm -it.

BROAD INSTITUTE CONFIGURATION

Broad Institute project members can load all project secrets from Google Secrets Manager (via gcloud) and boot the portal directly by using the bin/load_env_secrets.sh script, thereby skipping all of the configuration/registration steps in DEPLOYING A PRIVATE INSTANCE

bin/load_env_secrets.sh -p (path/to/service/account.json) -s (path/to/portal/config) -r (path/to/readonly/service/account.json) -e (environment)

This script takes up to 7 parameters:

  1. SCP_CONFIG_NAME (passed with -p): Name of configuration JSON inside GSM.
  2. DEFAULT_SA_KEYFILE (passed with -s): Name of default service account keyfile inside GSM.
  3. READ_ONLY_SA_KEYFILE (passed with -r): Name of read-only service account keyfile inside GSM.
  4. COMMAND (passed with -c): Command to execute after loading GSM secrets. Defaults to 'bin/boot_docker'
  5. PASSENGER_APP_ENV (passed with -e; optional): Environment to boot portal in. Defaults to 'development'.
  6. DOCKER VERSION TAG (passed with -v): Sets the version tag of the portal Docker image to run. Defaults to 'development'
  7. PORTAL_NAMESPACE (passed with -n): Sets the value for the default Terra billing project. Defaults to single-cell-portal-development

The script requires two command line utilities: gcloud and jq. Please refer to their respective sites for installation instructions.

DOCKER RUN COMMAND ENVIRONMENT VARIABLES

There are several variables that need to be passed to the Docker container in order to run properly:

  1. CONTAINER_NAME (passed with --name): This names your container to whatever you want. This is useful when linking containers.
  2. PROJECT_DIR (passed with -v): This mounts your local working directory inside the Docker container. Makes doing local development via hot deployment possible.
  3. PASSENGER_APP_ENV (passed with -e): The Rails environment you wish to load. Can be either development, test, or production (default is development).
  4. MONGO_LOCALHOST (passed with -e): Hostname of your MongoDB server. This can be a named host (if using a service) or a public IP address that will accept traffic on port 27107
  5. MONGO_INTERNAL_IP (passed with -e): Internal IP address of your MongoDB server. This is for the Ingest Pipeline to use when connecting to MongoDB if your firewalls are configured to only allow traffic from certain IP ranges.
  6. SENDGRID_USERNAME (passed with -e): The username associated with a Sendgrid account (for sending emails).
  7. SENDGRID_PASSWORD (passed with -e): The password associated with a Sendgrid account (for sending emails).
  8. SECRET_KEY_BASE (passed with -e): Sets the Rails SECRET_KEY_BASE environment variable, used mostly by Devise in authentication for cookies.
  9. SERVICE_ACCOUNT_KEY (passed with -e): Sets the SERVICE_ACCOUNT_KEY environment variable, used for making authenticated API calls to Terra & GCP.
  10. READ_ONLY_SERVICE_ACCOUNT_KEY (passed with -e): Sets the READ_ONLY_SERVICE_ACCOUNT_KEY environment variable, used for making authenticated API calls to GCP for streaming assets to the browser.
  11. OAUTH_CLIENT_ID (passed with -e): Sets the OAUTH_CLIENT_ID environment variable, used for Google OAuth2 integration.
  12. OAUTH_CLIENT_SECRET (passed with -e): Sets the OAUTH_CLIENT_SECRET environment variable, used for Google OAuth2
  13. SENTRY_DSN (passed with -e): Sets the SENTRY_DSN environment variable, for error reporting to Sentry integration.
  14. GA_TRACKING_ID (passed with -e): Sets the GA_TRACKING_ID environment variable for tracking usage via Google Analytics if you have created an app ID.
  15. PROD_DATABASE_PASSWORD (passed with -e, for production deployments only): Sets the prod database password for accessing the production database instance. Only needed when deploying the portal in production mode. See config/mongoid.yml for more configuration information regarding the production database.

RUN COMMAND IN DETAIL

The run command explained in its entirety:

INGEST PIPELINE AND NETWORK ENVIRONMENT VARIABLES

The Single Cell Portal handles ingesting data into MongoDB via an ingest pipeline that runs via the Google Genomics API (also known as the Pipelines API, or PAPI). This API should have been enabled as a previous step to this (see DEPLOYING A PRIVATE INSTANCE).

In order for the ingest pipeline to connect to MongoDB, and in addition to the variables relating to MongoDB as described in DOCKER RUN COMMAND ENVIRONMENT VARIABLES, there are two further environment variables that may need to be set. If you have deployed your MongoDB instance in GCP, and it is not on the "default" network, then you must set the following two environment variables to allow the ingest pipeline to connect to MongoDB:

Both of these pieces of information will be available in your VM's information panel in GCE. To set these variables, as there are not associated flags for bin/boot_docker, please export these two environment variables before calling any boot scripts for the portal:

export GCP_NETWORK_NAME=(name of network)
export GCP_SUB_NETWORK_NAME=(name of sub-network)

bin/boot_docker (rest of arguments)

This will set the necessary values inside the container. If your database is on the default network, this step is not necessary. If you are using a third-party MongoDB provider, please refer to their documentation regarding network access.

Single Cell Portal team members should set the associated values inside their GSM configurations for their instance of the portal.

ADMIN USER ACCOUNTS

The Single Cell Portal has the concept of a 'super-admin' user account, which will allow portal admins to view & edit any studies in the portal for QA purposes, as well as receive certain admin-related emails. This can only be enabled manually through the console.

To create an admin user account:

DEVELOPING OUTSIDE THE CONTAINER

Developing on SCP without a Docker container, while less robust, opens up some faster development paradigms, including live css/js reloading, faster build times, and byebug debugging in rails. See Non-containerized development README for instructions

HYBRID DOCKER LOCAL DEVELOPMENT

SCP team members can develop locally using Docker with a 'hybrid' setup that merges the speed of developing outside the container with the ease of using Docker for package/gem management. This allows using vite for hot module replacement (HMR) along with all the other features of native Rails development, but uses docker-compose to build and deploy containers locally. Both containers are built off of the single-cell-portal:development Docker image referenced above.

This setup will behave almost exactly like developing outside of the container - you can edit files in your IDE and have them updated in the container in real time, and any JS/CSS changes will automatically reload thanks to HMR without any page refresh. The startup process (minus any docker pull latency) should only take ~30s, though if you update any JS dependencies it can take 2-3 minutes to run yarn install on boot depending on how constrained your local Docker installation is in terms of CPU/RAM. You can reduce this latency by rebuilding the single-cell-portal:development Docker image locally. See the section on building the Docker image above for more information.

USING DOCKER COMPOSE LOCALLY

To leverage this setup, there are two shell scripts for setup and cleanup, respectively. To start the local instance, run:

bin/docker-compose-setup.sh

This will pull all necessary secrets from Google Secrets Manager and write env files to pass to Docker, then create two containers locally: single_cell (application server), and single_cell_vite (vite dev server). Both containers will install their dependencies automatically if you have updated any gems or JS packages and then start their required services. In addition to the Rails server, the single_cell container will run any required migrations and start Delayed::Job on startup. Both containers can be stopped by pressing Ctrl-C. This will not remove either container, and they can both be restarted with the same command.

You can also start these containers headlessly by passing -d. This runs the same startup commands, but will not attach STDOUT from the containers to the terminal. When using this option, note that it will take between 30-60 seconds before the portal instance is available at https://localhost:3000/single_cell.

bin/docker-compose-setup.sh -d

To stop both services and clean up the Docker environment, run:

bin/docker-compose-cleanup.sh

This will remove both containers, their dependent volumes, and reset any local configuration files to their non-Dockerized format.

UPDATING JAVASCRIPT PEER/INDIRECT DEPENDENCIES

Since yarn does not automatically upgrade peer/indirect dependencies automatically (it will only update packages declared in package.json), The following snippet can be used to update dependencies not covered in yarn upgrade:

https://gist.github.com/pftg/fa8fe4ca2bb4638fbd19324376487f42

CACHING IN LOCAL DEVELOPMENT

Single Cell Portal employs action caching for all data visualization requests, such as viewing cluster, violin, or dot plots. As such, making code changes during local development may not always reflect the most recent changes as a particular request may have already been cached by the server.

In order to turn caching on/off dynamically without having to restart the server, developers can run the following command from the terminal:

bin/rails dev:cache

This will add/remove a file in the tmp directory called caching-dev.txt that will govern whether or not action caching is enabled. NOTE: if you are developing inside Docker, you must run this command inside the running container.

If you want to clear the entire cache for the portal while running, you can enter the Rails console and run the Rails.cache.clear command which will clear the entire Rails cache (actions and assets, such as CSS and JS assets):

root@localhost:/home/app/webapp# bin/rails c
Loading development environment (Rails 5.2.4.4)
2.6.5 :001 > Rails.cache.clear

This will then return a list of cache paths that have been deleted. If you encounter a Directory not empty error when running this command, simply retry until it succeeds, as it usually only takes one or two attempts.

TESTS

UNIT & INTEGRATION TESTS

There is a unit & integration test framework for the Single Cell Portal that is run using the built-in test Rails harness, which uses Test::Unit and minitest-rails. These tests can be run natively in Ruby once all project secrets have been loaded. Each test handles its own setup/teardown, and no test data needs to be seeded in order to run tests.

To run all unit/integration tests:

bin/rails test

To run a specific unit/integration test suite (like the search API integration suite):

bin/rails test test/api/search_controller_test.rb

To run individual tests in a give suite:

bin/rails test test/api/search_controller_test.rb -n /facet/ # run all tests with the word 'facet' in the name
bin/rails test test/api/search_controller_test.rb -n 'test_should_search_facet_filters' # run exact match on name

Additionally, these tests can be run automatically in Docker. To run the unit & integration test suite, the boot_docker script can be used:

bin/boot_docker -e test -k (service account key path) -K (read-only service account path)

This will boot a new instance of the portal in test mode and run all associated tests, not including the UI test suite.

If you are a Broad engineer developing the canonical Single Cell Portal, run unit and integration tests like so:

bin/load_env_secrets.sh -e test -p scp-config-json -s default-sa-keyfile -r read-only-sa-keyfile

It is also possible to run individual tests suites by passing the following parameters to boot_docker. To run all tests in a single suite:

bin/boot_docker -e test -k (service account key path) -K (read-only service account path) -t (relative/path/to/tests.rb)

For instance, to run the FireCloudClient integration test suite:

bin/boot_docker -e test -t test/integration/fire_cloud_client_test.rb (... rest of parameters)

You can also only run a subset of matching tests by passing -R /regexp/. The following example would run all tests with the word 'workspace' in their test name:

bin/boot_docker -e test -t test/integration/fire_cloud_client_test.rb -R /workspace/ (... rest of parameters)

It is also possible to pass a fully-qualified single-quoted name to run only a single test. The following example would run only the test called 'test_workspaces' in test/integration/fire_cloud_client_test.rb

bin/boot_docker -e test -t test/integration/fire_cloud_client_test.rb -R 'test_workspaces' (... rest of parameters)

GOOGLE DEPLOYMENT

PRODUCTION

The official production Single Cell Portal is deployed in Google Cloud Platform. Only Broad Institute Single Cell Portal team members have administrative access to this instance. If you are a collaborator and require access, please email scp-support@broadinstitute.zendesk.com.

For Single Cell Portal staff: please refer to the SCP playbook and associated Github actions on how to deploy to production.

NON-BROAD PRODUCTION DEPLOYMENTS

If you are deploying your own production instance in a different project, the following VM/OS configurations are recommended:

For more information on formatting and mounting additional persistent disks to a GCP VM, please read the GCP Documentation.

To deploy or access an instance of Single Cell Portal in production mode:

PRODUCTION DOCKER COMMANDS

When you launch a new instance of the portal, you should get a response that is looks like a long hexadecimal string - this is the instance ID of the new container. Once the container is running, you can connect to it with the docker exec command and perform various Rails-specific actions, like:

STAGING

There is also a staging instance of the Single Cell Portal used for testing new functionality in a production-like setting. The staging instance URL is https://single-cell-staging.broadinstitute.org/single_cell

The run command for staging is identical to that of production, with the exception of passing -e staging as the environment, and any differing values for hostnames/client secrets/passwords as needed.

For Single Cell Portal staff: please refer to the SCP playbook and Jenkins server on how to deploy to staging.

Note: This instance is usually turned off to save on compute costs, so there is no expectation that it is up at any given time

TERRA INTEGRATION

The Single Cell Portal stores uploaded study data files in Terra workspaces, which in turn store data in GCP buckets. This is all managed through a GCP service account which in turn owns all portal workspaces and manages them on behalf of portal users. All portal-related workspaces are within the single-cell-portal namespace (or what the PORTAL_NAMESPACE environment variable is set to), which should be noted is a separate project from the one the portal itself operates out of.

When a study is created through the portal, a call is made to the Terra API to provision a workspace and set the ACL to allow owner access to the user who created the study, and read/write access to any designated shares. Every Terra workspace comes with a GCP storage bucket, which is where all uploaded files are deposited. No ACLs are set on individual files as all permissions are inherited from the workspace itself. Files are first uploaded temporarily locally to the portal (so that they can be parsed if needed) and then sent to the workspace bucket in the background after uploading and parsing have completed.

If a user has not signed up for a Terra account, they will receive and invitation email from Terra asking them to complete their registration. While they will be able to interact with their study/data through the portal without completing their registration, they will not be able to load their Terra workspace or access the associated GCP bucket until they have done so.

Deleting a study will also delete the associated workspace, unless the user specifies that they want the workspace to be persisted. New studies can also be initialized from an existing workspace (specified by the user during creation) which will synchronize all files and permissions.

OTHER FEATURES

ADMIN CONTROL PANEL, DOWNLOAD QUOTAS & ACCESS REVOCATION

All portal users are required to authenticate before downloading data as we implement daily per-user quotas. These are configurable through the admin control panel which can be accessed only by portal admin accounts (available through the profile menu or at /single_cell/admin).

There are currently 7 configuration actions:

READ-ONLY SERVICE ACCOUNT

The read-only service account is a GCP service account that is used to grant access to GCS objects from 'public' studies so that they can be rendered client-side for the user. Normally, a user would not have access to these objects directly in GCS as access is federated by the main service account on public studies (private studies have explicit grants, and therefore user access tokens are used). However, access tokens granted from the main service account would have project owner permissions, so they cannot be used safely in the client. Therefore, a read-only account is used in this instance for operational security.

This account is not required for main portal functionality to work, but certain visualizations (Ideogram.js & IGV.js) will not be enabled unless this account is enabled.

To enable the read-only service account:

  1. Create a new service account in your GCP project and grant it 'Storage Object Viewer' permission (see 'GCP Service Account keys' under DEPLOYING A PRIVATE INSTANCE for more information)

  2. Export the JSON credentials to a file and save inside the portal home directory.

  3. When booting your instance (via bin/boot_docker), make sure to pass -K (/path/to/readonly/credentials.json)

  4. Once the portal is running, navigate to the Admin Control Panel with an admin user account

  5. Select 'Manage Read-Only Service Account Terra Registration' from the 'Other Tasks' menu and click 'Execute'

  6. Fill out all form fields and submit.

  7. Once your new service account is registered, create a Config Option of the type 'Read-Only Access Control'

  8. Set the type of value to 'Boolean', and set the value to 'Yes', and save

This will have the effect of adding shares to all public studies in your instance of the portal for the read-only service account with 'View' permission. This will then enable the portal to stream certain kinds of files (BAM files, for instance) back to the client for visualization without the need for setting up and external proxy server.

To revoke this access, simply edit the configuration setting and set the value to 'No'.

TERRA ACCESS SETTINGS

Disabling all Terra access is achieved by revoking all access to studies directly in Terra and using the portal permission map (study ownership & shares) as a backup cache. This will prevent anyone from downloading data either through the portal or directly from the workspaces themselves. This will have the side effect of disallowing any edits to studies while in effect, so this feature should only be used as a last resort to curtail runaway downloads. While access is disabled, only the portal service account will have access to workspaces.

Disabling compute access will set all user access permissions to READER, thus disabling computes.

Disabling local access does not alter Terra permissions, but prevents users from accessing the 'My studies' page and uploading data through the portal. Downloads are still enabled, and normal front-end actions are unaffected.

Re-enabling Terra access will restore all permissions back to their original state.

SYNTHETIC DATA

Synthetic study data for testing and demonstrating portal features is configured in db/seed/synthetic_studies. See the Synthetic study README

MAINTENANCE MODE

The production Single Cell portal has a defined maintenance window every Thursday from 12-2PM EST. To minimize user dispruption when doing updates during that window (or hot fixes any other time) the portal has a 'maintenance mode' feature that will return a 503 and redirect all incoming traffic to a static maintenance HTML page.

To use this feature, run the bin/enable_maintenance.sh [on/off] script accordingly.

RUBOCOP

This project is configured to use RuboCop for static code analysis. There is a shell script that allows for running RuboCop only on files that have been edited since the last commit to Git. Configuration for RuboCop can be found here, or referenced from RuboCop's default configuration.

USAGE

To run RuboCop normally on all your local changes:

bin/run_rubocop.sh

Run in "lint-only" mode (does not enforce style checks):

bin/run_rubocop.sh -l

Run in "safe auto-correct" mode (will automatically correct any issues, where possible):

bin/run_rubocop.sh -a

The -a and -l flags can be used together. Usage text for the script can be printed with -h.