Create manual_deployment.md

docs/development/manual_deployment.md

@@ -0,0 +1,92 @@
+# Manual deployment process
+
+This document is useful to understand the low-level deployment process.
+
+For higher level overview, see [/docs/development/releases-and-deployment.md](/docs/development/releases-and-deployment.md)
+
+### Create the deployment package
+
+* ```npm run build --file=<OUTPUT_FILE_NAME>```
+  * We just tar.gz the source directory while excluding some files.
+  * might add some more stuff in the future - basically we want as much work done here as possible so deployment is quick
+
+### Upload the deployment package to slack
+
+We use slack as an easy to use repository of deployment packages.
+
+* ```npm run upload --file=<PACKAGE_FILE_NAME> --token=<SLACK_API_TOKEN>```
+  * creates a file called <PACKAGE_FILE_NAME>.slack_url containing the private package url
+
+### Download a package from the private package url
+
+* ```npm run download --file=<OUTPUT_FILE_NAME> --url=<PACKAGE_PRIVATE_URL> --token=<SLACK_API_TOKEN>```
+  * downloads from the given private slack url to the given output file name
+
+### Deploy the package
+
+To deploy a package file:
+
+* extract the package (tar.gz)
+* create a .env file in the package directory with relevant configurations
+* run the deploy command from the root of the extracted deployment pacakge
+  * ```npm run deploy```
+
+## Testing the deployment procedures using vagrant
+
+There is a vagrant machine which allows to test the deployment flow locally.
+
+It runs the flow of building a package, then extracting and deploying it locally inside the vagrant instance.
+
+You can also download a package from slack (using ```npm run download```) and then deploy it.
+
+To use it, [install vagrant](https://www.vagrantup.com/docs/installation/), then run the following from spark/vagrant directory:
+
+```
+spark$ cd vagrant
+spark/vagrant$ vagrant up
+```
+
+Now, you can ssh into the instance, the deployed package is at /opt/spark/deployment
+
+When you run node / npm - you will have correct version for the deployment.
+
+For example, to start the spark web-app:
+
+```
+spark/vagrant$ vagrant ssh
+ubuntu@ubuntu-xenial:~$ cd /opt/spark/deployment
+ubuntu@ubuntu-xenial:/opt/spark/deployment$ node server.js
+```
+
+## Automated deployment using Travis-CI
+
+Travis is run on every pull requests / commit to branch.
+
+Have a look at the [.travis.yml](/.travis.yml) and [/.travis/on_build_complete.sh](/.travis/on_build_complete.sh) to see what it does.
+
+### Travis build flow
+
+* runs the tests
+* on failure - just notify to slack (using ```npm run log```)
+* if tests are successfull and it's not a pull request:
+  * builds the package (using ```npm run build```)
+  * uploads the package to slack (using ```npm run upload```)
+  * sends notification to slack containing the built package url (using ```npm run log```)
+
+### Testing travis build locally
+
+You should set the following environment variables (modify accordingly..)
+
+```
+SLACK_API_TOKEN=""
+SLACK_LOG_WEBHOOK=""
+TRAVIS_PULL_REQUEST="false"
+TRAVIS_REPO_SLUG="Midburn/Spark"
+TRAVIS_BRANCH="master"
+TRAVIS_BUILD_NUMBER="5"
+TRAVIS_BUILD_ID="198736627323"
+SPARK_DEPLOYMENT_KEY=""
+SPARK_DEPLOYMENT_HOST=""
+```
+
+Now, you can run the commands from .travis.yml file