Skip to content

Latest commit

 

History

History

scripts

README.md

Scripts to Deploy of JetBrains Projector on Google Cloud Run

The scripts in this folder automate the deployment described in our JetBrains Projector documentation.

Before using the scripts, please configure the "projector" by commenting and uncommenting the proper variables in .env.

Control Scripts

The following scripts are used to deploy and delete projector images in Cloud Run:

Deployment Aspects

Keep Containers Active When Idle

Problem: Auto-Shut-Down of Idle Projector Containers

By default, CloudRun shuts idle projector containers down after 15 minutes. As a consequence, users have to re-configure Rider.

Solution: Disable Auto-Shut-Down in CloudRun Configuration

Deploying containers using the flags --no-cpu-throttling and --min-instances 1 we prevent the automatic shut down.

Details about configuring the CloudRun deployment are explained in the associated CPU allocation (services) documentation.

Caveat: Changing the CloudRun Service Configuration Results in Re-Deployment

Caveat: Changing the configuration of a CloudRun deployment results in a re-deployment. After a configuration change, e.g. by using the enable-... and disable... scripts, you have a fresh container instance and need to configure Rider and the project again.

Test the Default Deployment and Enable/Disable-... Scripts

Links for the checks required in the Gherkin THEN paragraphs below:

SCENARIO 1: Default Deployment with Always-On
GIVEN normal deployment of 2 projector-rider instances 
AND the projector configuration is completed for projector-rider-2
AND > 15 minutes passed without interaction

WHEN I reload the projector-rider-2 browser page using the same URL

THEN the state of projector-rider-2 is the same as before the reload,
i.e. I don't need to configure the IDE again
AND the service configuration page for projector-rider-2 shows:
    `CPU is always allocated`, `Minimum number of instances 1`, `Maximum number of instances 1`
AND metrics show that the `projector-rider-2` has been billable for the entire period of time
SCENARIO 2: Enable Auto-Shutdown
GIVEN normal deployment of 2 projector-rider instances
WHEN I execute the enable-... script
AND I complete the projector configuration for projector-rider-2
AND I wait for > 15 minutes without interaction
AND I reload the projector-rider-2 browser page using the same URL

THEN the state of projector-rider-2 differs from before the reload,
i.e. I need to configure the IDE again
AND the service configuration page for projector-rider-2 shows:
    `CPU is only allocated during request processing`, `Minimum number of instances 0`, `Maximum number of instances 1`
AND metrics show that the `projector-rider-2` has NOT been billable for the entire period of time
SCENARIO 3: Disable Auto-Shutdown
GIVEN normal deployment of 2 projector-rider instances
AND I have executed the enable-... script

WHEN I execute the disable-... script
AND I complete the projector configuration for projector-rider-2
AND I wait for > 15 minutes without interaction
AND I reload the projector-rider-2 browser page using the same URL

THEN the state of projector-rider-2 is the same as before the reload,
i.e. I don't need to configure the IDE again
AND the service configuration page for projector-rider-2 shows:
    `CPU is always allocated`, `Minimum number of instances 1`, `Maximum number of instances 1`
AND metrics show that the `projector-rider-2` has been billable for the entire period of time

Note:

The specifications explicitly take metrics from projector-rider-2 in order to prevent an error in looping through all configured instances. I have not considered edge cases in the tests (i.e. test the last instance, test the first instance, etc.), because of time and quite low risk - assuming that we will not change the code structure of the scripts!

Helper Scripts

The following script is called by the control scripts described above: