GitLab Runner UX, 4 May 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Runner Fleet: CI job queues vision

Description

Gina Doyle (Product Designer, Runner & Pipeline Insights) walks through the vision for Runner Fleet CI job queues. Note: This vision is not set in stone, it will constantly evolve and has no timeline associated with it.

Please leave comments & feedback in the vision issue: https://gitlab.com/gitlab-org/gitlab/-/issues/345594/

All other resources can be found in the issue above ^

A

Hi everyone, I'm gina doyle and I'm the product designer for runner and pipeline insights today I'll be walking you through our new ci job q's vision for the runner fleet category we've heard a lot of requests from customers around this, so we're really excited to share it before I get into it. I want to note that this is a vision, so it's not set. It will evolve as we get feedback on it and we don't have a timeline associated with it just yet.

A

So, let's start with what we've heard from customers. So if you can imagine yourself as a software engineer and you're running pipelines, there is little information today that is available to see how soon a ci job may start.

A

You may be left asking why your job is pending for so long as a platform engineer using the admin runners view or the group runners view, you're then responsible for investigating why the jobs aren't running, there's nothing to help you determine if the runner is busy today, which would mean determining the number of running jobs. A runner has the number of pending jobs in its queue, as well as an estimated time of when they will be picked up. So without this information, it's difficult to help your team get back to their daily work.

A

We have taken a stab at solving this problem in the admin runners area through these visionary designs, so we'll be focusing on solving it for priyanka. First, like I said, these will continue to evolve so they're not set in stone, but I will hop over into figma, so I can show you the prototype.

A

Okay, so just as you see the admin runners area today, you'll come to a runner list view you'll see a new job, cues tab up top. That gives you insight into which runners or runner tags have pending jobs, what their average q duration is now and trends over time and how to optimize your infrastructure to save costs and time.

A

So, let's check it out once you get to the page, you'll see a list of runners with queued jobs, you'll get a status of the latest job that they're running and then the ability to filter by that that um runner or view the runner details you'll also see the average queue duration for all of your instance runners, because right now we're only viewing instance, runners and some q duration trends over the last three months.

A

Finally, we'll provide you with some recommendations of how to optimize your infrastructure, for example, by looking into the runners with the highest q durations in this instance.

A

So if I do, filter by runner filtering by runner will give you only that runner's queued jobs.

A

So that's what this list is: you'll get the position number for each of those jobs in the queued, duration or the time it will take for that job to be picked up. You'd also be able to remove a job from the queue and, for example, sign assign it to another runner or just completely trash it again, we'll provide you with recommendations based on your trends over time and your average q duration. So, for example, in this one, we noted your average q duration for this runner is increasing.

A

So you probably would want to create another runner with the same architecture to be able to service the jobs quicker you'll also be able to interact with the graph as well and see how trends um the queue trends over time and there's different colors, representing how the duration compares to the threshold. So, for example, the dotted line is the threshold which is actually the average and then, as your q duration goes up, you'll see that it becomes red and um just indicating that it's over that threshold.

A

So if there is a time that you're looking at runner tags, instead of specific runners, you'll also be able to filter by those. So if I were to clear this filter, um I would be able to get some recommendations on which runners to filter by because we know how many runners have this many pending jobs. But if I know the tags that I want to filter by, I can just input those in so the filter would work, just as um it does in the admin area today.

A

Filter for this time sake, I did it quickly and now you'll see all the cue jobs for my runners that have the tags that are noted here with that same information that I was given before, but it's specific to this set of runners like the average queue duration, q, duration, trends, etc.

A

um So that's it for the prototype and I'm going to jump back over to this. I will be linking the vision issue in this video and, as we continue to iterate this, we would really appreciate your feedback. So we'd love for you to add some comments into the issue. If you have any feedback which, once again I will be linking and that's about it, thank you.