Kong Kong Summit 2021, 2 Nov 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Multi-Cloud Connectivity with a Service Mesh ZeroLB Pattern

Description

In this demo, Ross McDonald shows how Kong Mesh delivers secure, seamless application connectivity across different clouds and runtimes, allowing administrators to abstract multi-cloud and on-premise infrastructure into a truly “write once, run anywhere” model.

▬▬▬▬▬▬ TIMECODES ▬▬▬▬▬▬

0:00 - Intro
0:10 - Zero-Load Balancer (ZeroLB) Architecture Overview
1:28 - Redis Counter Application Without Locality-Aware Load Balancing
2:00 - Kong Mesh Setup Overview
2:41 - Enabling Locality-Aware Load Balancing in Kong Mesh
4:03 - Simulating Success
4:30 - Simulating Failure
5:18 - Conclusion

▬▬▬▬▬▬ LINKS ▬▬▬▬▬▬
❏ Learn more about the ZeroLB Pattern: https://konghq.com/blog/zerolb/
❏ Watch more recordings from Kong Summit here: https://konghq.com/kong-summit/2021-videos

▬▬▬▬▬▬ COMMUNITY ▬▬▬▬▬▬

⭐ https://github.com/Kong/kong
🐦 https://twitter.com/thekonginc
🌎 https://konghq.com/community/
🍻 https://www.meetup.com/pro/kong
❓ ️https://discuss.konghq.com/
💯 https://konghq.com/kong-champions/

#ServiceMesh #LoadBalancing #KongMesh #KongSummit

A

Hey there, and welcome. for this demo,, I'm going to walk you through a multi-cloud, multi-zone.

A

Setup of kong mesh. before we dive in,. Let's take a quick look at the architecture. here,. We have a.

A

Single service mesh joined across three separate cloud, deployments,, otherwise known as a multi-zone.

A

Deployment. two are in gcp,, east and west,, and one is in aws east.. Each cluster is fronted.

A

With a kong api gateway, where traffic is routed to our front end application,, which then connects to.

A

Redis to retrieve a counter., all of the services across every region are deployed in a service.

A

Mesh., so each pod in each cluster has a sidecar proxy that talks to the rest of the services in.

A

The mesh. it's through these sidecar proxies where a mutual tls is enforced and load balancing takes.

A

Place. during the demo,, we will showcase how kong mesh provides seamless application connectivity.

A

Across availability, zones,, regions, and clouds all while providing critical functionality to.

A

Maintain the health of our applications., we'll start by showing the counter app and.

A

How request to redis are automatically load balanced across the three available backends.

A

We will then dynamically change the load, balancing policy to enable locality, awareness,.

A

Which tells the mesh to prioritize communication to services that are nearest geographically.

A

We'll then, simulate a failure to show how the mesh automatically reroutes traffic to other regions to.

A

Maintain uptime and availability, all without manual intervention. and with that,. Let's get.

A

To it. here is our counter application. here's the counter,. It is being retrieved from redis, as well.

A

As the zone in which it was fetched, from. you'll notice that if I refresh the page here a few times.

A

That the returned zone varies.. This is because we don't currently have locality, aware load, balancing.

A

Enabled within the mesh., so even though I've loaded the counter app from the east ingress.

A

It may be fetching results from amazon, from gke-east, gke-west.. This is introducing extra.

A

Latency in our application and reducing overall service performance. to fix this,.

A

We'll want to enable the locality aware load, balancing policy within the mesh.. Here's the.

A

Kong mesh control plane,, which provides a high level overview of the zones, services, proxies,.

A

And policies that are being enforced within the service mesh. from here,. I can see the three zones.

A

That we have deployed, one in amazon-us-east,, one in gcp-east, and another gcp-west.

A

I can also see the services that we have deployed,, including our counter application and redis. each.

A

Have three data plane proxies,, signifying that one proxy is available in each region.

A

And I can even drill down into those proxies here from the ui and see metadata associated.

A

With each one., in order to enable locality aware load, balancing, we'll need to update the mesh.

A

Custom resource definition that lives within the control plane. to do that,. I will need a terminal.

A

We can see what the current mesh definition looks like by issuing a get mesh, command. here,.

A

You can see we have an empty routing component. in order to enable locality, awareness, we'll need.

A

To issue a patch to add an extra value here. to do this,, I'm going to issue a kubectl patch command.

A

I'm going to add a localityawareloadbalancing, and it's going to be set to true. before I do. This,.

A

I'm going to enable the auto increment feature inside of the counter app., with.

A

Auto increment enabled,, the counter will be automatically incremented, and we can see that the zone is currently swapping between east, west, and aws..

A

Now,, if I issue this patch command,.

A

The counter is now pinned to the gke-east zone.. I can then even look at my west ingress.

A

And validate the same is taking place for the west ingress where west is pinned to.

A

West and east is now pinned to east.. I can go back to our control plane,.

A

The default mesh, and you'll note now that locality, aware load balancing is.

A

Now set to enabled., if I go back to the counter application and increment, you'll note that now.

A

We've stabilized on the east zone, redis., so east ingress,. Now it goes to the gke-east,.

A

And I'll click, this a few more times, and you'll see this holds. now,. If I load the west ingress.

A

And do the same,, the north of the zone for the gke-west is now pinned to gke-west redis.

A

Now, let's simulate a failure to see what happens with our routing. to do this, I'll pull up another.

A

Terminal here. I'll need to switch clusters to my west cluster here. and then from here,.

A

I'm actually going to scale down my redis deployment,, where I'm using kubectl, again,.

A

Scale deployment redis in my main application, namespace to 0 replicas, effectively, shutting.

A

Off redis in the west, gke environment. you'll note that it's still working. and if I scale this down,.

A

So I've scaled it down, hit increment here, again.

A

Note that now the mesh is automatically routed. The requests around the west instance.

A

And is now back to the previous load, balancing behavior between gke-east and amazon-us-east.

A

All of this without having to change any configuration or modify anything manually.

A

In this demonstration,, we showcased how kong mesh provides robust and resilient.

A

Connectivity to services and applications deployed across multiple clouds and regions.