Cloud Native Computing Foundation KubeCon + CloudNativeCon North America 2018 (Seattle), 15 Dec 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Lightning Talk: Kubegres: Accessing K8s From Postgres - Liz Frost, Heptio

Description

Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io

Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects

Lightning Talk: Kubegres: Accessing K8s From Postgres - Liz Frost, Heptio

Postgres is a very versatile database. In addition to regular SQL queries, you can use a Foreign Data Wrapper to make anything - other databases, JSON files, Twitter - appear to be just another set of tables in the database. This includes Kubernetes! In this lighting talk I will show off querying cluster information, starting and stopping pods, and more… all from inside a postgres SQL prompt. Along the way, I’ll explain a bin about the Postgres foreign data wrapper model, how to use the Go client for Kubernetes, and show off some fun statistics you can gather when your whole cluster is a database.

To Learn More: https://sched.co/GrVY

A

I forgot what I submitted for the description? I can't do half of those things, hello, everybody! Thank you for staying till the end, my name is Liz, frost I'm still in beta on Twitter, and if you want to find me on the Kate slack, I am just Liz. I am a queer trans and polyamorous woman. I am Co mom to a very cute and very fuzzy dog.

A

If you follow me on Twitter you're, mostly going to get pictures of flurry, I work for a company called hep do which, as of right now, is still an independent company but is soon to be a part of VMware and perhaps most importantly, to all of y'all I created the caverna T's cube, cuddle, Pony stickers are still available. Come find me afterwards. I'm not here, to talk about any of that, though I'm here to talk about Postgres, specifically I'm, going to talk about Postgres foreign data rappers get get it tough crowd.

A

Foreign data rappers are away for Postgres to query and manage data not stored in the database itself. They can be mundane like ODBC or sequel Lite, something more exotic like Cassandra or Redis, or even weirder. There are FTW s4 like hue bulbs, there's a list of FD W's, and you can use in FTW to query that list and now I'm gonna, add Q Burnet E's to the mix. Here's my very advanced technical diagram of how this is going to work.

A

In essence, we'll put a caverna, T's client inside your database and have it talk to the Taverna DS API server, I'm a convert, Indies contributor, so unfortunately I'm mostly familiar with go Postgres rappers are written in C, but thankfully somebody did the hard work of bridging the gap for me already shoutout to den WC on github I'm gonna gloss over the days of debugging seg faults, because this is a five-minute talk and I'm gonna skip ahead to where it's actually working.

A

The first step is that we install the library we do this by calling create extension inside Postgres, which tells it to load the object file, and then we can create a server which, for us just means a cluster I, pass into a cube config, so that we can actually access it, and then we create a table. This is pretty freeform. You can have as many or as few fields as you'd like I'm, starting off with name and names face. The alias option tells our rapper, which fields to retrieve from the communities object for those columns.

A

We use the options field at the bottom to say what kind of objects we're looking for in this case we're looking at pods and we're going to use the the gate system namespace, because I couldn't be bothered to make an example. So this is just cube a DM and then we run a query. This is a five minute talk, so you're not going to get a live demo from me. You're just gonna have to believe me when I say this actually works, but don't be sad, I've simulated demo syndrome.

A

For you, this error keeps me up at night because it's something very gnarly to do with memory corruption, but it does go to show that we are actually making Network calls. If you can fix this, please come find me. I have no idea. If, at first you don't succeed, try again consistently running the same query again on a new connection, basically always works, so you can see our names and our namespaces of a newly created tates master node with cubed IAM right there. But what about something more complicated?

A

We've got containers in a pod, but there's more than one container per pod. What if we just want to pull the first one out, each pod can have multiple containers, so we can use JSON path. If you don't know JSON path, it's probably one of the most useful parts of the cube cuddle the command, not the pony. It provides a useful, possibly familiar syntax for us. Let's add that to our table.

A

We use the same, alias option we used before, but this time we wrap it in curly braces to signal its JSON path and it works. The container URLs show up exactly as we expect them to lists are all fine and dandy, but how about something? A little less structured labels are a good candidate for this, because they're freeform, but still important. Fortunately, Postgres- has a built-in json type.

A

That'll work perfectly for this, so we have that in and here we go a bunch of JSON and to prove to you that that isn't just a text field, we can query individual fields on the labels. I think that's pretty cool. It's like a little MongoDB inside your sequel and for my last trick, I'm running out of time, but I've got one last thing: I'm going to need a few more tables. This is way too much text for a slide, but fortunately we don't care about most of it.

A

We're just going to create new replica sets and deploy mint tables, and that's right, we're gonna do joins. We can use the fact that pod names, replica names and deployment names share, prefixes and combine them together wearing know more about getting your pods and deployments confused. This is absolutely not production, ready software error handling when the columns don't matches. Try not to do that.

A

If you do get the right tall, M type and it's not one of the ones I added, specifically, it's still going to be very rough and the code base is well I wrote it it three days ago, there's more unit tests than you'd expect and more C than you'd, probably like. Thank you very much for your time. We're on this is on github. This is on slack and enjoy the rest of your evening. Come fine, I'm, pretty distinctive.