GitLab CI/CD, 7 Jan 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: GitLab 13.6 Feature Flag Webhook and GitLab 13.7 Auto Rollback new features

Description

Short description and demo of Feature Flag Webhook feature introduced in GitLab 13.6 and Auto Rollback feature introduced in GitLab 13.7

Some useful links:
Cesar Saavedra's Twitter: @cesar_saavedr
Link to the repository containing this demo:
https://gitlab.com/tech-marketing/sandbox/cd/prodmgr-spring
Link to the documentation for Feature Flag Webhook:
https://docs.gitlab.com/ee/user/project/integrations/webhooks.html#feature-flag-events
Example webhook receiver:
https://docs.gitlab.com/ee/user/project/integrations/webhooks.html#example-webhook-receiver
Cloud-based webhook receiver:
https://webhook.site/
Link to the documentation for Auto Rollback:
https://docs.gitlab.com/ee/ci/environments/index.html#auto-rollback
Our Progressive Delivery direction:
https://about.gitlab.com/direction/ops/#progressive-delivery
Post-deployment monitoring Epic:
https://gitlab.com/groups/gitlab-org/-/epics/3088

A

Hello, my name is sister cevader, I'm a technical marketing manager at gitlab in this segment of the presentation, I'm going to be covering a new capability called a fire web hook on feature flag change.

A

This feature was introduced in 13.6 and uh and it consists or basically uh generated an event when there is a feature, flag, activation or deactivation this this event uh in in turn it invokes a newly uh this newly introduced webhook and the webhook receiver can then take some action as a result of this invocation.

A

Why does it matter for customer and prospects as a developer or as someone that wants to monitor the enabling enabling and disable disabling of a feature flag? You can use this new feature the this new web hook.

A

You know as when, using this feature you can take some action as a result of uh the invocation of this web hook. uh You could, for example, kick off a cicd pipeline. You can send a message to slack or to any other chat ops platform. You can send an email or text. You can also update and process. um You know some application performance monitoring solution that you may have and- and you can also do business activity, monitoring, updates and processing as well.

A

This capability uh streams lines the proc processes that need to take a feature flag being updated, uh enable or disable in this case into account here are some resources uh to learn more about this feature. This is a link to the documentation of the issue, some things to follow a couple of links there about testing web hooks and troubleshoot web hooks from the gitlab platform, there's also a sample for a webhook webhook receiver that you can install on your laptop and a link to a cloud-based webhook receiver.

A

That requires no installation, which is what I used for this, for this example on how to show you uh how it works so step one you can set up your web hook and I'm using the site called webhook.site for this, and you set up a webhook there you get the url, then you go to gitlab settings webhooks window and you basically paste that url and that's step three and also in the in in that screen.

A

You need to scroll down and make sure you select feature flags, events select that check box and then you click on, add a web hook and that will add the web hook so that when a feature flag is enabled or disabled, that webhook is invoked and the step four to test it out. uh If you already have an existing feature flag in this case, you can enable it and disable it, and then you will see the web hook being called as a result of that event.

A

All right, so the next uh capability, I'd like to go over is called automatic rollback in case of failure. This was introduced in gitlab 13.7.

A

And it's basically uh the automatic rollback in case a critical alert is detected, so this is done. uh You know by the gitlab platform and what gets rolled back is the most uh recent uh successful deployment. um It is selected and redeployed.

A

So why does it matter for customers and prospects? This capability allows sres and operators to set up critical alerts in production so that, when those alerts are triggered in a problematic deployment, let's say is you know a threshold for the amount of memory that your application is using?

A

Gitlab will roll it back to the most recent successful deployment. This automation eases the described workflow above for sres and operators. It also serves like a second line of defense for code updates that pass ci tests and are deployed to production.

A

Here are some resources to learn more about this specific capability, there's a link for the documentation and issue, and I will paste a link to the demo once it's ready. Some things to follow. There is a link to the progressive delivery direction that talks about where we're heading with the these type of type of capabilities, and also the post deployment monitoring epic. That contains a list of all the related work items.

A

You know that are within progressive delivery and include this capability and capabilities like this one, too things to keep in mind. The rollback is skipped if a deployment is running when the alert is detected and also a rollback can happen only once in three minutes.

A

If there are multiple alerts uh detected at once, only one rollback is performed.

A

Okay. Now, let's move on to the demo.

A

So here you see a project called pro manager spring and I'm logged on as devin. He is a devops engineer and um first thing to we need to do. Is we need to enable.

A

Automatic deployment rollbacks so, as you can see here, it's already enabled and what this will do is if there's a critical alert triggered the rollback.

A

There will be an automatic rollback done by uh by the platform to the last successful deployment.

A

So that's already set.

A

Well, one thing: one more thing I want to show: you is what I'm going to be deploying um by starting this pipeline in a minute introduces some changes and the changes are here.

A

In the index.html file, the headings of the table are all in capital, letters and there's also a misspelling here.

A

So those will be the changes that are in those are the changes in master and the ones that I will be pushing by creating or running a new pipeline.

A

A

Here and as you can see, there's been two successful deployments earlier today. Please ignore this cancel once I started them by accident and now we're going to run a.

A

Pipeline and while that's running, let's go to the operations environment actually, metrics.

A

And create an alert, so let's create an alert here. Let's make sure that this is the total memory usage and, let's make sure, to create an alert that will be triggered.

A

So, let's make uh let's see the average is 2.61 so and right now it's 2.45. So let's make it to.

A

So two right there.

A

Okay, so it detected that it's below the threshold, so it should fire an alert.

A

That hasn't fired yet.

A

Let's see okay, so the pipeline is running okay, so now that the pipeline has completed uh the deployment has taken place and let's see if the alerts has come in.

A

Yep there's a critical alert that already happened a few minutes ago. So, let's see, if that triggered a.

A

A new deployment.

A

So, let's go to environments um now notice here that the job that was executed is actually not the job in the pipeline. So if you look at the pipeline, let's go to the pipeline, so the pipeline that executed.

A

Here seven minutes ago, it finished it has an automatic uh deployment to production. This is not a manual step, okay. Now. What I would like you to notice here is that this was updated eight minutes ago right, so it was um updated after this pipeline completed, which means that the the critical alert triggered an automatic uh rollback and also notice that the job is a job that does not belong to the pipeline that we just ran right.

A

This is, in fact, let's go there, and this job is actually a job of the last successful uh the last successful pipeline.

A

So the automatic rollback has taken.

A

A

Yeah, so the latest successful deployment included the edit and the new product are have a purple background, but if you notice the titles do not have, they are not in capital. Letters like we uh we saw earlier, which is what we pushed in the in the pipeline that actually was automatically rolled back.

A

That's all I have thank you very much.