Skip to content

Latest commit

 

History

History
105 lines (80 loc) · 4.98 KB

README.md

File metadata and controls

105 lines (80 loc) · 4.98 KB
Raw

Running the export-answers script

This script finds all student answers in a Firetore report-service database, collects all the answers for a single user's assignment into a single Parquet file, and uploads all the files to S3 for querying by the Athena database.

// build the shared library
cd functions
npm install
npm run build:shared

// build the scripts
cd scripts
npm install

Copy the config.json.sample file to config.json and fill in the AWS credentials for the S3 bucket. This should be available in 1Password as one of the report-service AWS accounts.

Download the credentials.json file for the appropriate Firebase project from e.g. https://console.cloud.google.com/apis/credentials?project=report-service-pro&organizationId=264971417743

Run the script by passing in the destination S3 bucket and the source key for the Firestore database. e.g.

node export-answers.js concord-staging-report-data authoring.staging.concord.org

the script accepts up to three arguments:

export-answers.js [s3-bucket] [source-key] [created-timestamp-regex]

  • s3-bucket: the name of the s3 bucket we're going to put the answers in

  • source-key: the Firestore answers collection

  • created-timestamp-regex: (optional) a regex as a string, e.g. 2021-(04|05)

    if this is included we will gain a slight speed-up by not uploading answers whose created property does not match the regex. Note that this is conservative: if an answer does not have a created field it will be uploaded, and it only takes one answer in a user's session that matches the regex (or has created missing) to upload the entire learner-assignment collection.

Running on Lightsail

In order to run the script on production, it is useful to run it on a Lightsail server, which should keep running for as long as needed, and may be slightly faster.

  1. Log into AWS Lightsail
  2. Select Linux/Unix and a Node.js blueprint
  3. Select a price tier. For the entire DB dump we used one of the more expensive options ($40/mo, over 2 days =~ < $3), but for smaller dumps smaller tiers may be sufficient
  4. Name the instance e.g. export-answers-to-s3 and create it
  5. Select the instance once it has booted up (~5 mins) and take note of the ssh username and ip
  6. To connect to it, you need a valid private key, which you can download from https://lightsail.aws.amazon.com/ls/webapp/account/keys, and you could save in e.g. ~/.aws/LightsailDefaultKey-us-east-1.pem (and chmod 600)
  7. ssh into the server with ssh -i ~/.aws/LightsailDefaultKey-us-east-1.pem [username]@[ip]
  8. git clone the repo git clone https://github.com/concord-consortium/report-service.git && cd report-service
  9. npm install and build the shared scripts 
    cd functions
npm install
npm run build:shared
cd ../scripts
npm install
  10. In a different terminal window, from the /scripts folder, scp the credentials and config files 
    scp -i ~/.aws/LightsailDefaultKey-us-east-1.pem config.json [username]@[ip]:~/report-service/scripts
scp -i ~/.aws/LightsailDefaultKey-us-east-1.pem credentials.json [username]@[ip]:~/report-service/scripts
  11. Use screen to start a session that won't be terminated when you log out 
    screen        # start a new screen session
ctrl-a d      # exit the session (without terminating it)
screen -ls    # view running screens
screen -r [processname]   # return to a screen
  12. Run the node export-answers.js as in the section above

Running on Google Cloud Virtual Machine

It can also be useful to offload the running of scripts to a virtual machine in Google Cloud. They will usually run faster there.

  1. Log in to Google Cloud at https://cloud.google.com/ and go to the Console.

  2. Click the Compute Engine option, then click "Create Instance".

  3. Select the Marketplace option and search for "nodejs". Select the Free filter, then choose Node.js - Google Click to Deploy, and click the Launch button.

  4. Name the instance, set the Zone value to match the Firestore instance's zone (currently us-central1 for report-service-pro), and then set Series to "E2" and Machine Type to "e2-standard-8", then click the Deploy button.

  5. With the VM running, SSH into it. You can SSH via a web browser window from the VM console by selecting SSH > Open in browser window.

  6. Once you're connected, git clone the repo: git clone https://github.com/concord-consortium/report-service.git && cd report-service

  7. npm install and build the shared scripts

    cd functions
npm install
npm run build:shared
cd ../scripts
npm install

  8. Manually add the credentials.json file to the /scripts folder so you can access Firebase. If you'll be using a script that interacts with an AWS S3 bucket (e.g., export-answers.js), also copy the config.json.sample file in the scripts folder to config.json and fill in the AWS credentials for the S3 bucket. Credential details for Firebase and AWS are available in 1Password.

  9. Run scripts, e.g., node export-answers.js