GitLab Engineering Productivity Showcases, 14 Mar 2022

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: Triage Ops: Why and How to Dry Run

Description

Dry runs help identify that the expected actions of a scheduled automation happen on the first run. This can be useful to identify large amounts of notifications that may be triggered from comments and that the number of updated records matches expectations

Dry run output for the example in the video is at https://gitlab.com/gitlab-org/quality/triage-ops/-/jobs/2200795513

A

Hey this is kyle, um a member of the team was asking questions about, dry, runs for scheduled automation and triage ops, and I was about to run one. So I thought I would just record a quick video to give an overview.

A

First off dry runs are helpful because they allow you to see what the scheduled automation is going to do, based on the current conditions over what over the scope that is run at so dry runs can be run over groups or specific projects, just like all gitlab triage policies and I'm going to open one and then I'm going to trigger one here shortly.

A

So this policy, I added a new rule to remove this label when severity is applied, so we have a a rule that adds automation, ux, missing labels when there's no severity for sus issues, and it leaves a comment that says: please remove it manually and I thought why. Why require people to remove it manually? We can do that automatically.

A

So, let's add a condition where, when issues have that label and have severity, we remove it and we do that silently so no comment, so there's no notification in emails and things like that.

A

um So what I did is I identified: what is the scope of issues that I would have expected at the group level to be affected based on the different severity levels where it well? Let me just open this up where the automation ux, missing labels label is present and there's severity, one for example. So here you can see two and in this case, since there is a severity, we want to go ahead and remove that label.

A

I noticed that when I added everything up, I thought 249 issues and I had already run a dry run, we'll talk through the results, but one of the things that jumped out there's only 191 issues identified, so it didn't match the scope that I expected. So that was one thing that I looked for in this output. The other thing that I looked for was what action was going to be taken on a specific issue.

A

So here you can see this kind of pattern repeated over and over again, where there's a link to an issue and then inside of triple backticks the action that would be taken based on this policy and in this case this issue, you can kind of see here that it has automation, ux, missing labels and severity 4, so that is going to be unlabeled or removed from the from that issue.

A

So I went ahead and made a change. uh There used to be a condition here. That said, only run this on open issues that was the source of the difference between 2 and 49 and 191.

A

So now I'm going to run another dry run. So if I look at the most recent pipeline here, I'm going to open the dry run custom job and what we're going to do? Apologies for jumping around so much here.

A

Is we're going to add three different variables that allow us to control which rule is run and over what scope so.

A

Okay, so here this section describes how to test with the dry run and it mentions you need to have the triage policy file, triage source type and triage source path, so we start with triad source type. This is the variable that gitlab triage, the the gem that we are using for triage ops, scheduled automation, says: do I want to do projects or groups so we're going to do groups and there's not much forgiveness here? If you say group, it is just groups, so don't forget the f. um Then the next is triage source path.

A

So this is where you would define the project, and you can do this with a string or an. I, though, well I'll, say the group. You can do this with a string, but I prefer to use the id, which is also accepted so go to the gitlab.org group copy the id over time.

A

You may memorize that it's 9970 and then lastly, we will do the triage policy file here, I'm going to copy the relative path of the file from the mr, then we're going to trigger it and we'll give it some time to run um actually. No we're not um I'll I'll just end the video here and I'll link to the driver and output in the description. Thanks all hope. This was helpful.