Generic Container Timeout support

[eak13]

As laid out in #533, timeouts are currently ignored by the executors. To begin provide this support, this issue asks the following:

1. Update the Airshiptctl API interface to add a variable to pass a timeout value (likely seconds).
2. Update the the Generic Container framework accept the timeout variable & then inject it into the container to prompt the operation to finish gracefully. At this point in time it will not force kill the process.

[teoyaomiqui]

From the proposed airship design discussion, here are key directions:

airshipctl container executor implementation has to pass the timeout to container implementation:

• Executor implementation
• Container API interface

There are two ways to do so:

1. Pass timeout as an environment variable to the container implementation as parameters to Run() function. And then inject timeout as ENV variable into the actual container to make it aware of the timeout.

• The main benefit of this patchset is that if in the future we would want to kill the container by airshipctl forcefully, we can do so almost effortlessly.
• The downside would be that we need to change Container API interface.

2. Inject the timeout as an environment variable into the container interface by extending ENV variable slice similarly to the kubeconfig context variable here.

• The benefit of this approach will not require any changes to Container API interface.
• The downside would be that if we decide to delete the container forcefully by airshipctl, we will have to extract this environment variable ourselves somehow.

Timeout needs to be respected by the containers that we run.

We use 3 main images for those containers currently:

1. Toolbox

• Container executor YAML definition examples

• Image source code

• Dockerfile

  We must modify toolbox scripts to support the timeout

2. Cloud-init container for the ephemeral host

• Container executor YAML definition

• Image source code

• Dockerfile

  This is an intermediate container that is used only to prepare files for the image-builder.

  We should remove this container and instead build a wrapper around the image-builder container that would allow us to execute it without the need to have two different phases, one for cloud-init, another one for image-builder

3. Image-builder

• Container executor YAML definition

• Image source code and Dockerfiles

  We need to investigate if the image-builder container respects timeout and implement support if it's not.

[vladiskuz]

Can someone assign it to me?

[airshipbot]

Related Change #792801

Subject: Pass TIMEOUT_SECONDS env variable to generic container API
Link: https://review.opendev.org/c/airship/airshipctl/+/792801
Status: NEW
Owner: Vladislav Kuzmin (vkuzmin@mirantis.com)

Approvals

Code-Review
! None
Verified
+1 Zuul
+1 ATT Airship2.0 CI
Workflow
! None

Last Updated: 2021-05-28 05:39:55 CDT

[vladiskuz]

Move cloud-init KRM to image-builder https://review.opendev.org/c/airship/image-builder/+/802046

[joshuaherrera]

It looks like @vladiskuz may be working on something else currently, I can take on finishing this as I have capacity.

[eak13]

@joshuaherrera added you to this issue

[vladiskuz]

@joshuaherrera
This issue is almost done. I need to change a minor comment here https://review.opendev.org/c/airship/airshipctl/+/792801
Also we have another patch set for container timeout https://review.opendev.org/c/airship/airshipctl/+/805619

[joshuaherrera]

@vladiskuz Since it is close to completion, I will unassign myself so you can finish this issue. Apologies.

[michaelfix]

@vladiskuz , @joshuaherrera , Relates to: #597 is mentioned in the commit message for the PS above - https://review.opendev.org/c/airship/airshipctl/+/805619. Is 805619 related to this 544 issue here, or 597?

[gtsteffaniak]

please assign to me

[raliev12]

Per Slack discussion, please assign this issue to me for finalizing.