Today we’re happy to announce something really exciting: the introduction of fully managed Linkerd to Buoyant Cloud. This means that if you’re using Linkerd, Buoyant Cloud will soon be able to automate your Linkerd upgrades, installations, rollbacks, trust anchor rotations, data plane version sync, and more, right on your own clusters. Combined with the existing comprehensive health monitoring and alerting for all things Linkerd, this is a massive step towards our goal of giving you a zero-stress service mesh.
Managed Linkerd is currently in private beta for a select few Buoyant Cloud customers. If you’d like to join the pilot program, please sign up for the waitlist!
From the very beginning,1 Linkerd has been designed for simplicity. It’s especially designed for operational simplicity: the vital but all-too-ignored characteristic of giving operators a system that’s understandable and predictable, that doesn’t involve complex tuning, leaky abstractions, surprises, or unknown unknowns.
As anyone who’s tried to design distributed systems knows, building simplicity is anything but simple. The world of Kubernetes is notorious for its complexity. And while we don’t believe this reputation is 100% earned, the service mesh landscape is even more notorious for complexity. In these murky, bug-infested waters, Linkerd stands out, high and dry: it can deliver even complex features like mutual TLS without exposing that complexity to the user, and without the burden of a complex, leaky proxy layer. Linkerd is completely unique in the service mesh world.
But as simple as Linkerd is, the fact is that running software—any software—still, well, sucks. Software needs to be upgraded; software needs to be monitored; software needs to be cared for.
At Buoyant, we’ve heard this same thing loud and clear from our many Linkerd adopters. As amazingly simple as Linkerd is, tasks like upgrades and trust anchor rotation are unpleasant, boring, and even a little scary. In other words: they’re toil, the bane of SREs everywhere.
With fully managed Linkerd, we’re taking a big step towards fixing that problem.
Not really, of course! Ha ha. Well, probably not. But we are automating away all the stress and toil related to running Linkerd. We’re doing this in two ways:
The end result? You can sleep happy at night and be productive during the day.
Of course, things can still go wrong. Let’s not pretend they can’t. Linkerd sits at the intersection of many sources of failure, including the network, the cluster, and, of course, your application itself. But we’re going to do this in a way that is predictable and that gives you ultimate control. For example, keeping your data plane in sync with your control plane requires—thanks to Kubernetes’s immutable pods—restarting your workloads. Buoyant Cloud will do that on only the workloads you allow, and in the exact way you specify, without surprises. Upgrading your control plane will happen only when you are ready, and if anything goes tragically wrong, the existing control plane will not be removed. And so on.
We love service meshes and we love geeking out about them, so it might sound a little surprising when we say our goal is to make Linkerd boring. But it is. This is not an easy job. It’s profoundly difficult to make something simple out of complex components, and Linkerd, by virtue of the substrate of Kubernetes below it if nothing else, is built from complex components.
But simple, and boring, is the goal, and pretty much everything we’re doing in Buoyant Cloud is designed to achieve that goal.
Give it a try, and let’s see just how boring we can make it.
(And if you’re attending ServiceMeshCon in Valencia later this month, join me for my keynote: An update on the extremely boring and uninteresting world of Linkerd)
Over the next few months, we’ll be rolling out managed Linkerd to our Buoyant Cloud customers. Want a sneak peek at managed Linkerd now? Let’s chat!