# Create a remote Zulip dev server This guide is for mentors who want to help create remote Zulip dev servers for hackathon, GCI, or sprint participants. The machines (droplets) have been generously provided by [DigitalOcean](https://www.digitalocean.com/) to help Zulip contributors get up and running as easily as possible. Thank you DigitalOcean! The `create.py` create uses the DigitalOcean API to quickly create new virtual machines (droplets) with the Zulip dev server already configured. ## Step 1: Join Zulip DigitalOcean team We have created a team on DigitalOcean for Zulip mentors. Ask Rishi or Tim to be added. You need access to the team so you can create your DigitalOcean API token. ## Step 2: Create your DigitalOcean API token Once you've been added to the Zulip team, [log in](https://cloud.digitalocean.com/droplets) to the DigitalOcean control panel and [create your personal API token][do-create-api-token]. **Make sure you create your API token under the Zulip team.** (It should look something like [this][image-zulip-team]). Copy the API token and store it somewhere safe. You'll need it in the next step. ## Step 3: Configure create.py In `tools/droplets/` there is a sample configuration file `conf.ini-template`. Copy this file to `conf.ini`: ``` $ cd tools/droplets/ $ cp conf.ini-template conf.ini ``` Now edit the file and replace `APITOKEN` with the personal API token you generated earlier. ``` [digitalocean] api_token = APITOKEN ``` Now you're ready to use the script. ## Usage `create.py` takes two arguments - GitHub username - Tags (Optional argument) ``` $ python3 create.py $ python3 create.py --tags $ python3 create.py --tags ``` Assigning tags to droplets like `GCI` can be later useful for listing all the droplets created during GCI. [Tags](https://www.digitalocean.com/community/tutorials/how-to-tag-digitalocean-droplets) may contain letters, numbers, colons, dashes, and underscores. You'll need to run this from the Zulip development environment (e.g. in Vagrant). The script will also stop if a droplet has already been created for the user. If you want to recreate a droplet for a user you can pass the `--recreate` flag. ``` $ python3 create.py --recreate ``` This will destroy the old droplet and create a new droplet for the user. In order for the script to work, the GitHub user must have: - forked the [zulip/zulip][zulip-zulip] repository, and - created an SSH key pair and added it to their GitHub account. (Share [this link][how-to-request] with students if they need to do these steps.) The script will stop if it can't find the user's fork or SSH keys. Once the droplet is created, you will see something similar to this message: ``` Your remote Zulip dev server has been created! - Connect to your server by running `ssh zulipdev@.zulipdev.org` on the command line (Terminal for macOS and Linux, Bash for Git on Windows). - There is no password; your account is configured to use your SSH keys. - Once you log in, you should see `(zulip-py3-venv) ~$`. - To start the dev server, `cd zulip` and then run `./tools/run-dev.py`. - While the dev server is running, you can see the Zulip server in your browser at http://.zulipdev.org:9991. See [Developing remotely](https://zulip.readthedocs.io/en/latest/development/remote.html) for tips on using the remote dev instance and [Git & GitHub Guide](https://zulip.readthedocs.io/en/latest/git/index.html) to learn how to use Git with Zulip. ``` Copy and paste this message to the user via Zulip chat. Be sure to CC the user so they are notified. [do-create-api-token]: https://www.digitalocean.com/community/tutorials/how-to-use-the-digitalocean-api-v2#how-to-generate-a-personal-access-token [image-zulip-team]: http://cdn.subfictional.com/dropshare/Screen-Shot-2016-11-28-10-53-24-X86JYrrOzu.png [zulip-zulip]: https://github.com/zulip/zulip [python-digitalocean]: https://github.com/koalalorenzo/python-digitalocean [how-to-request]: https://zulip.readthedocs.io/en/latest/development/request-remote.html ## Updating the base image Rough steps: 1. Get the `ssh` key for `base.zulipdev.org` from Tim, Vishnu or Rishi. 1. Power up the `base.zulipdev.org` droplet from the digitalocean UI. You probably have to be logged in in the Zulip organization view, rather than via your personal account. 1. `ssh zulipdev@base.zulipdev.org` 1. `git pull upstream main` 1. `tools/provision` 1. `git clean -f`, in case things were added/removed from `.gitignore`. 1. `tools/run-dev.py`, let it run to completion, and then Ctrl-C (to clear out anything in the RabbitMQ queue, load messages, etc). 1. `tools/run-dev.py`, and check that `base.zulipdev.org:9991` is up and running. 1. `> ~/.bash_history && history -c && sudo shutdown -h now` to clear any command line history (To reduce chance of confusing new contributors in case you made a typo) and shut down the droplet. 1. Go to the Images tab on DigitalOcean, and "Take a Snapshot". 1. Wait for several minutes. 1. Do something like `curl -X GET -H "Content-Type: application/json" -u : "https://api.digitalocean.com/v2/images?page=11" | grep --color=always base.zulipdev.org` (maybe with a different page number, and replace your API_KEY). 1. Replace `template_id` in `create.py` in this directory with the appropriate `id`. 1. Test that everything works. 1. Open a PR with the updated template_id in zulip/zulip! ## Creating a new base image Creating a new base image happens rarely since updating the base image is good enough most of the time. Check out https://chat.zulip.org/#narrow/stream/3-backend/topic/new.20base.20dev.20droplet to view the discussion when we attempted to do upgrade last time. Rough steps: 1. Get the `ssh` key for `base.zulipdev.org` from Tim, Vishnu or Rishi. 1. Log in to the Zulip organization view, rather than via your personal account. 1. Create a new droplet in DigitalOcean with 2GB RAM and `base.zulipdev.org` as the SSH authentication key. 1. Log in to the droplet as root user. Make sure to point the SSH program to the private key of `base.zulipdev.org` during this step. 1. Create a user called `zulipdev` and add it to the `sudo` group. 1. Make sudo of `zulipdev` user passwordless by including `zulipdev ALL=(ALL) NOPASSWD:ALL` in `/etc/sudoers.d/90-cloud-init-users` 1. Copy the `authorized_keys` file of `root` user to the `.ssh` directory of `zulipdev` user 1. Switch to `zulipdev` user and set the permissions for the `.ssh` folder to `700` and `.ssh/authorized_keys` to `600`. 1. Clone `https://github.com/zulip/zulip` repository in the home directory of `zulipdev`. 1. `git remote rename origin upstream` 1. `git clean -f`, in case things were added/removed from `.gitignore`. 1. `tools/provision` 1. Insert `NODENAME=zulip@localhost` to `/etc/rabbitmq/rabbitmq-env.conf` 1. `tools/provision` 1. `tools/run-dev.py`, and check that `base.zulipdev.org:9991` is up and running. 1. Clone `https://github.com/zulip/python-zulip-api` repository in the home directory of `zulipdev`. 1. `git remote rename origin upstream` 1. `git clean -f`, in case things were added/removed from `.gitignore`. 1. `./tools/provision`. 1. `> ~/.bash_history && history -c && exit` 1. SSH to root user. 1. `> ~/.bash_history && history -c && sudo shutdown -h now` 1. Go to the Images tab on DigitalOcean, and "Take a Snapshot". 1. Wait for several minutes. 1. Do something like `curl -X GET -H "Content-Type: application/json" -u : "https://api.digitalocean.com/v2/images?page=11" | grep --color=always base.zulipdev.org` (maybe with a different page number, and replace your API_KEY). 1. Replace `template_id` in `create.py` in this directory with the appropriate `id`. 1. Test that everything works. 1. Open a PR with the updated template_id in zulip/zulip! ## Remotely debugging a droplet To SSH into a droplet, first make sure you have a SSH key associated with your GitHub account, then ask the student to run the following in their VM: ``` $ python3 ~/zulip/tools/droplets/add_mentor.py ``` You should now be able to connect to it using: ``` $ ssh zulipdev@.zulipdev.org ``` They can remove your SSH keys by running: ``` $ python3 ~/zulip/tools/droplets/add_mentor.py --remove ``` # Creating a production droplet `create.py` can also create a production droplet quickly for testing purposes. ``` $ python3 create.py --production ```