wrata

This project is used to run scripts on distributed network with parallel execution

How it works

Basically workflow consists of several simple steps:

1. Creating DigitalOcean droplets (Server class model in project code) with specified count. Count of servers depends of tasks. Some projects has a small number of rspec files, so 1-5 droplets is enough Some of them got 10000s of specs and requires 10s of droplets
2. Adding rspec files from project to test list with specified parameters (testing server url, language, browser, etc...)
3. Each of those rspec files are run on one of droplets, chosen from free one in queue
4. After each of rspec files are finished - servers can be turn-off manually or via non-active timeout (Default 1 hour)

How to start debug session

1. Start docker container with disposable database:

   docker run --rm \
     --name wrata-debug-psql \
     -e POSTGRES_HOST_AUTH_METHOD=trust \
     -e POSTGRES_USER=postgres \
     -e POSTGRES_DB=wrata_development \
     -d \
     -p 5432:5432 \
     postgres

2. Run DB migrations:

   ENV=develoment rake db:create db:migrate db:seed

3. Start Rails debug session in RubyMine. Set RAILS_MASTER_KEY env correctly.

4. Open http://localhost:3000 in browser and create new admin with username from config/credentials.yml.enc

5. Do your debug

6. After debug is done stop DB. It automatically removed after stop:

   docker stop wrata-debug-psql

How to update

This should be done after ~1 hour after merging PR to master Because this task will create docker hub image with name onlyoffice/wrata:latest and this task is not very fast

git pull --prune
docker compose pull app
docker compose up -d

Credential options

This config is stored in config/credentials.yml.enc file, encrypted
To edit stored data you should create file config/master.key with secret word specified secure password storage (search password file by config/master.key)
And after that call bin/rails credentials:edit to edit secret file, to add or remove values

Example of file layout is:

secret_key_base: key # secret key
github_user: user # user of github to fetch projects
github_password: pass # password of user
ssh_user: user # username for node
ssh_pass: pass # password for node
admin_emails:
  - email1@domain1.com
  - email2@domain.com

SSL Setup

By default this repo require SSL certificates to work.
./certs/tls.crt and ./certs/tls.key should exists on your host, outside docker
You can disable it by changing nginx.conf

How to update nginx container

docker-compose stop
docker-compose rm web
docker-compose pull web
docker-compose up -d

How to backup database

docker-compose exec db pg_dumpall -c -U postgres | gzip > dump_$(date +"%Y-%m-%d_%H_%M_%S").gz

How to restore database backup

docker-compose stop app
zcat dump*.gz | docker exec -i testing-wrata_db_1 psql -U postgres
docker-compose start app

Admin account

Admin account is account that have same name as email in admin_emails list in secrets Admin account is different in two ways:

1. Admin can add new users to wrata via https://wrata-url/clients page
2. Admin email receive notifications about with some failure information which includes:
   • spec_no_tests_executed_email - this mean some rspec do not output any result. Usually this mean something is not properly configure. For example spec contains not a single it
   • spec_failed_email - this mean spec failed mid-execution. Usually this is a networking error and simple restart of spec can help
   • All other failures - this is some bad code. Sometimes then creating Droplet in DigitalOcean - DO servers return 500 instead of creating ones. Usually fixed by restarting create process

Troubleshooting

• For uninitialized constant HtmlWithPassedTime (NameError) error in logs you should add rspec_passed_time_formatter gem dependency in project file.
  Since #857
• High SQL storage usage - this is normal, because we store all logs in database and they are not deleted automatically.
  You can delete them manually by logging as active user, going to Profile -> Run history -> Clear history.
  After that you should wait some time for postgres to delete all data or trigger VACUUM manually
• node is hangup on creation
  Usually this happens when DigitalOcean return 500 error on server creation.
  It's out of our control, you only can recover in this situation
  First via DigitalOcean control panel check status of this server (is it created or not)
  Usually it will be already created (if not - just change wrata status to destroyed) Next open https://wrata-url/servers, find the node in question and change it's status to 'created'
  After that usually self-destruction mechanism triggered due to known problem, and server will be destroyed automatically
  After that recreate server as usual, via Create button or REST API

How to release new version (for maintainers)

1. Update VERSION file
2. Update CHANGELOG.md by adding version line after master (unreleased)
3. Create PR with those changes and merge to master
4. On master run rake add_repo_tag
5. On GitHub create new release via web-browser and add info from CHANGELOG.md