Cloud Native Computing Foundation KubeCon + CloudNativeCon North America 2016 (Seattle), 16 Nov 2016

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Lightning Talk - The Feature Formally Known as Prince ... err PetSet.

Description

Lightning Talk - The Feature Formally Known as Prince ... err PetSet. Stateful Sets in 600 Seconds or Less - Chris Love, Datapipe

What, When and Why to use a PetSet.

About Chris Love
Chris has been coding for over 20 years and is an open source contributor for both Kubernetes and Terraform. He has architected key projects with such companies as Accenture, Motorola, ADP, Inuit, Warner Bros, and CenturyLink. Chris is passionate about scalable open source technologies, Big Data, and not getting woke up at 2 am because a system went boom. Outside of work, Chris can be found working on his computer too much, enjoying time with wife, or passing on his love for the Martial Arts through Kenpo.
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.

A

Who's, tired yeah, all right. We got five minutes here, guys and I've got a live demo and we'll see. I already have one cloud fail on me today: I won't throw the vendor underneath the bus and we also have the second cloud. That was a backup.

A

It's working, coop got will get NL works and, let's see if our macbook is happy I'm going to have to flip screens and go over to the SharePoint hey and you know what, while he's doing this I can I can steal time so H a data distributed containers- oh my god, you know, can I, please buy a vowel vanna H. A data stores are typically within containers, a train wreck.

A

So let me start the demo, because I got five minutes and I need to deploy Sandra cluster yeah. All right. Anybody actually recognizes the song. You can smile in the purple shirt all right excellent. Now, if the gentleman has to come back and let's see if we can do this- and yes all right. So this is the feature formerly known as pet Prince pet set uh who's using pet sets right now. It's an alpha. Oh cool I get to talk about it. It's in alpha currently and we've.

A

The community as a whole is decided upon beta to change the name to stateful set I'm, actually not going to discuss that today. We're going to talk about why we use them.

A

I'll come on guys, there's a rough crowd tonight all right. So let's talk about some things that are really important within data world stable network at identity- oh my god, my product still uses IP addresses not D&S. Oh my god, I'm running zookeeper and I need oertel indexes means zookeeper one needs to start up and then two and then we got to do master election and yeah I won't cuss on the stage um stable storage.

A

Well, we're talking about distributed data stores across regions, so I need storage and then we're talking about the startup and teardown sequence, which is extremely important in terms of liveliness probes and readiness probes within the sequence now whose heads ready to explode anybody, you guys got to admit it I just like said 16 things in 30 seconds.

A

But that's not why I had the pitcher. Let's talk about the complications of doing stateless versus stateful, what does that mean to everybody everybody's doing microservices nowadays and they're going to save the world I'm old as an old dog I know what so a stack was so we're looking at typically HTTP stateless interfaces that can scale across your cluster.

A

But now we need to put the data somewhere. We've got to store the data. We've got to do upgrade, so we've got to do backups. Well. Can we get away from that? What happens when you go into a pet set model of a highly distributed data store? Is you open yourself up to risk? If you don't have to do it, don't do it? If you do have to do it, we have the tool.

A

Now this is going to be stable in beta, changing the name, and you know, frankly, it makes a life a little bit more complicated, but I think we'd be bored if it wasn't so I've got a minute. 40 left, let's go over here and let's do.

A

We got two notes grated it's taking an average three minutes to bring up the nodes. That's with attached, storage, Cassandra. You know pick your saiful data store as long as it plays well. You know there's a lot of complications within this and what we've trying to do, simplify that cluster is going to be running a couple more minutes, so evidently three nodes yeah. What about five minutes so basically, you're like at this point, Chris I've, been at this conference for two days I'm in information overload.

A

I gave you my handle at the beginning of the presentation, which is online I'm on I'm on slack way too much. So please reach out as well as take particular consideration to github com, kuber Nettie's, flat 4, /, charts I'll, give some credit to both google and Diaz for building a really cool tool. You prize guys have probably caught heard of already. It's called helm in the helm project because we're running alpha it's not as it's not moved into stable, so components such as zookeeper, etsy d, console Cassandra. You know all the regular suspects.

A

Are there eight seven? Thank you so much.