Kubernetes on Janusworx

Kubernetes & CNCF Awards

feedback@janusworx.com (Mario Jason Braganza) — Mon, 20 Apr 2026 11:15:00 +0530

Intended Audience

Mostly me.

Click the image to embiggen!

This is a note of gratitude.
A tad late. But better late than never :)

I decided to get involved with the Kubernetes community on a lark about four years ago.
I wanted to learn about containers and orchestration, and I thought learning via joining the upstream community was it.

I was wrong 😂
Life got in the way and I only finally got around to learning it all last December.

But the community was interesting and helpful.
I got to learn the terms of art, by osmosis.
I got to see how a large community functions, how a large software project was run.

And in time, I got to pitch in and help a bit as well.
I pitched in with the Release Team for v1.25 to gain a sense of how a release comes about.
And a couple of years ago, I saw my good friend (and frequent collaborator) Priyanka buckling a bit under the weight of the work, in the various roles she had assumed to help out in the community. I asked her if there was anyway I could help her, that did not involve meetings or talking to people 😂¹ And she asked me if I could pitch in and help with helping folks gain membership into various Kubernetes orgs on Github.

I ayed.

She taught me the ropes. And I became a Github New Membership Coordinator for the Kubernetes project.

And all of this history to say, I’m grateful to Priyanka for this serendipitous opportunity.
I fell in love with the work.
For someone who has had to face a lot of cynicism in his real world day job, it’s been a lot of joy, welcoming folks into the project. It’s one of the happier parts of my day, when I sit at the desk to process the requests that come in.

I’m also grateful to the Kubernetes as well as the overarching CNCF communities for recognising what I do, for two successive years.
I was the recipient of the Kubernetes Contributor award in 2024 and the CNCF’s, Chop Wood Carry Water award in 2025.
I’m both, honoured and humbled.
I hope to contribute to this community, a while longer :)

Feedback on this post?
Mail me at feedback at this domain.

P.S. Subscribe to my mailing list!

I was well and truly burned out then ↩︎

Kubernetes, Note to Self: Need Load Balancer Installed on Bare Metal

feedback@janusworx.com (Mario Jason Braganza) — Tue, 11 Nov 2025 07:37:54 +0530

Intended Audience

Mostly me. Also other grizzly sysadmins who are learning devops like me.

One thing that bit me when I was trying to expose my apps to the world when working on the home cluster, is that Kubernetes on bare metal—I was using Kind at the time—expects to talk to a load balancer service, which then talk to an actual load balancer. Which if you are using bare metal, you won’t usually have.
I had to then go expose a “NodePort” to gain access from outside.

So to expose my stuff in as “real world” a way as possible, we need to:

Either install a load balancer implementation like MetalLB. OR
Use a Kubernetes distribution that has a load balancer implementation built-in, like K3s.

I chose option 2 and used K3s, because I am, as they say in popular parlance, using Kubernetes at the edge.¹
In which case, I prefer to have as many batteries built-in as possible.

Feedback on this post?
Mail me at feedback at this domain or [continue the discourse here][2].

P.S. Subscribe to my mailing list!
Forward these posts and letters to your friends and get them to subscribe!

Although, from the articles I’ve read, if you’re doing a multiple node cluster, then you’re better off using MetalLB. ↩︎

K3s: Move Data Folder

feedback@janusworx.com (Mario Jason Braganza) — Mon, 10 Nov 2025 16:50:45 +0530

Intended Audience

Mostly me. Also other grizzly sysadmins who are learning devops like me.

Problem
#

My main SSD is too small to hold all my kubernetes and K3s shenanigans. (It’s only about 512gb) So I need a way to move my K3s data folder out to my big HDD.

Solution
#

Caution

Before you do any of the following, make sure to stop the cluster with:
sudo systemctl stop k3s

K3s supports having a custom data directory, by letting me specify a data-dir.
I decided to specify it in the K3s config file at /etc/rancher/k3s/config.
If it isn’t there (as it wasn’t in my case), just create one. You’ll need to do this as the root user.
This is what I now have in there.

data-dir: /BIGHDD/rancher/k3s

And then, I …

Created a rancher directory in my big hard disk. (root owns it)
Ran rsync as the root user: rsync -a /var/lib/rancher/k3s/ /BIGHDD/rancher/k3s/
Started up my cluster again with a sudo systemctl start k3s

The Plan for My New Hetzner VM

feedback@janusworx.com (Mario Jason Braganza) — Thu, 06 Nov 2025 08:30:54 +0530

Update: 2025-12-11
#

I got it done!

Preamble
#

If this post is weird and stream of consciousness like, that’s by design.
You are seeing a notes and miscellanea post, in the work category.
I want to be able to think through what I am doing. Mostly to clarify, document and explicitly state what I want. To avoid shifting goalposts (or rather to shift them intentionally, if need be.)
I’ll be updating and revising this post as I think and learn and experience things

Goals
#

Think about and write what I want to get done with my hetzner
For now, the how does not matter

What I want
#

A VM running K3s, that I can deploy my Hugo website to, as well all the apps that I choose to run on Janusworx
My main need with a cluster is the GitOps aspect. Not the scaling or redundancy. I want infrastructure as code and automated updates, as much as possible.
Everything goes through my Forgejo instance at home
I write a post and it should deploy to the website somehow
Apps that I put in there, either just my config files of already published images/deployments, or my own custom things, ought to be pushed to or pulled by the cluster and deployed. Upgrades should be a matter of manipulating code
Secrets should be self hosted, via environment variables or some self hosted solution. Never at a third party service

2025-11-06 09:35
#

Got tangled with Hugo in the morning.
I think I got the basics down, the VM comes up. K3s in installed, but in what order do I move my apps over?
Think I’ll the root domail last. And try beginning with the “simpler” apps first
Begin with familiarising myself with Helm
2025-11-06 10:06: Tea break
2025-11-06 13:06: Lunch break

2025-11-06 16:00
#

Back. Trying to port my dev terraform setup to production. It should just work. But it does not.
2025-11-06 16:59: Solved. I had somehow borked my Terraform state. Nuked it. And then things started working again
Show stopper now. My GPG key has expired. Need to figure out how to sign my git commits with SSH now
2025-11-06 18:26: I think I’ve licked it. Both Github and Forgejo don’t seem to be complaining. Let’s see what else I’ve broken over the coming days
2025-11-06 20:00: Done! My production cluster is up! And I can tear it down and build it up consistently. Now to work with it tomorrow.

Feedback on this post?
Mail me at feedback at this domain.

On How the Kubernetes v1.29 Logo Came About

feedback@janusworx.com (Mario Jason Braganza) — Fri, 15 Dec 2023 17:00:20 +0530

Kubernetes v1.29 was released a couple of days ago.
And just like every other release before it, it has a distinct name and theme (Mandala) as well as its own unique logo. In whose creation, I’m proud to say, I had a part to play!

The logo! Look, how pretty!

It all began when I helped the v1.29 Release Lead, mentee and trouble-maker in general, Priyanka Saggu set up her blog in 2018. I taught her all about CRAP¹ and I think all those crappy lessons lingered and made an impression.
So when it came time to design the release’s logo, she asked me if I could pitch in and help, because she was running short on time. It sounded like fun, so I said, Hell Yeah!

Partners in Crime aka Spitballing aka Brainstorming
#

Priyanka was very clear headed on her brief for what the logo was to be about.

Universal: We need to celebrate the fact that this project has folks all over the world, working on it.
Singularly Recognisable: It had to be recognised as the 1.29 logo at a glance at various scales

Everything else was up for grabs.
Thoughts meandered from parodying popular memes, like we do in our talks …

to cheeky play, with popular media like the old folks did, to seeking inspiration from music lyrics and nature patterns.
Priyanka then hit upon something else she wished to incorporate in the design. Could it have Indian elements? That led us down more fun alleys in the beautiful maze that is Indian Art. Across the length and breadth of our multicultural nation; we checked out everything …

from Warli art,

courtesy, Wikipedia

to Kalamkari

courtesy, Wikipedia

through Madhubani

courtesy, Wikipedia

before we settled on Mandala

courtesy, Sambeet D, Pixabay

Why Mandala?
#

Because it satisfied both the conditions of the brief, beautifully and succinctly.

Nearly every culture, from Hinduism to Buddhism to Mayan to Christianity, has some spin on the patterns and the symbols that make up the Mandala. It represents the cosmos and is used across the world. To our minds, nothing could be more universal than this art form
And because it is singularly recognisable.

Here it is in a slack chat, as an emoji²

It’s pretty distinctive! And as you look at it, at ever increasing sizes, more details emerge.
The midsized image I posted above is pretty detailed, but you can go watch it in all its uncompressed glory here³

Putting It All Together
#

The rest was then, figuring out the details.
Since one of us has quite limited design skills, that were rusty to boot⁴ and the other one had no time⁵, we decided to go look for a liberally licensed Mandala image, that we’d use as a base. We went through quite a few images, before we stumbled across Fibrel Ojalá’s beautiful creations.
And once that was done, it was just a matter of structuring it, the way we wanted it to look. We used Annapurna SIL⁶ for the text and Inkscape to put it all together.
The whole process took us a week, from start to finish.
It was both, a joy and a privilege to contribute to this release.

P.S. The Easter Egg
#

I thought it’d take a little longer than it did, but you folks are more eagle eyed than I thought possible!
Yes, that’s my little homage to Devanagari, the script lots of Indian langugages use for their writing. If you read the symbols in the diamonds clockwise, beginning at nine o’clock, they’re Devangari numerals that read 1.29 🙂

Feedback on this post? Mail me at feedback@janusworx.com

P.P.S. Subscribe to my mailing list!
Forward these posts and letters to your friends and get them to subscribe!
P.P.P.S. Feed my insatiable reading habit.

I know very little design, but what little I do know is burned into my psyche because of Robin Williams and her CRAPpy principles, I learnt in the late 90s ↩︎
apologies for the potato quality of the screenshot ↩︎
You’ll need Annapurna SIL for the text to render right ↩︎
Me ↩︎
Priyanka. She really can draw, though. I mean, have you seen her sketches? ↩︎
I saw it used in a hymnbook, a decade ago and loved it at first sight ↩︎

Learning Kubernetes, Setup Notes

feedback@janusworx.com (Mario Jason Braganza) — Thu, 03 Nov 2022 13:12:50 +0530

Ok, here’s to learning Kubernetes in earnest, to get that career pivot going.
Notes as I learn, over the coming days.

This one, is a lot of random thoughts on bringing up a cluster.

It is both harder than you think, and easier than you think
I did this, because I wanted to run my own cluster, just like the big boys do.¹
It’s much easier to run alternatives like kind, if that suits your needs better.

Notes to future Jason:

I made a gigantic Ubuntu KVM vm, for my playground
Then four small vms in that big one
Because of life, I goofed up and used Ubuntu LTS on the big vm and a non-LTS version on the small ones. Next time just use the same os everywhere.
If learning is your aim, make sure the vms use a beefy, full fat os install. The skinny ones might be helpful online when actually building stuff, but when it comes to learning, you’ll be installing a lot of things manually.²
To that end, make sure the virtual machines have lots of disk space available. I had to learn how to resize qcow files, because I nearly ran out of space.
The control-plane node will need two cpus. the others can get by with one.
Same with ram. Helps if the control-plane node has 4gb.
The documentation is the best resource. It is comprehensive but not well structured. No hand holding here.
I used containerd and runc rather than Docker, manually installing them using the documentation links³
Pair with a how to install guide, off the web. Kubernetes releases come fast and furious. Make sure you aren’t following something woefully out of date.
Once again, the documentation is the go to. Make sure you read them for notes and pitfalls. My cluster keeled and fell over after my first reboot, because I did not configure the cgroups driver right, to work with systemd
It takes time. It’s ok. It’s supposed to take that long.
If your system is old and creaky like mine, don’t bring up all the vms at once. That’s another surefire way of making sure the cluster goes splat. Bring up the control node, wait a while⁴, wait some more just to make sure, and then slowly bring up the worker nodes one by one.
Then keep the darned thing on all day.
If stuff works, and you like me, need to get on with life because you are doing all this in the middle of a house move, make sure to take lots of vm snapshots.
What else? Hmm … oh yea! Stop stressing. You got this. Show up. It’ll come to you.

Like Scott says, You got to do the real thing, if you want to learn well. ↩︎
I had to manually install apt-utils, whiptail, ping, devel tools and Dog knows, what else ↩︎
I honestly don’t why I did this. No objective reason. Just that the recent Docker emails give me the creeps. Reminds me of the old Microsoft emails. So, Via Negativa! ↩︎
Waiting time? Enough to brew me a coffee. Measure yours. ↩︎

Make KVM Use a NIC in Bridge Mode

feedback@janusworx.com (Mario Jason Braganza) — Tue, 18 Oct 2022 17:48:35 +0530

I wanted to start learning Kubernetes.
So I thought I’d spin up a Ubuntu VM to use as my sacrificial guinea pig.

While the normal NAT mode works well enough for most of my networking needs, I wanted the Ubuntu VM to show up and work as a machine on my host’s network.¹
I have two network cards on my desktop and wanted to allocate one of them to this VM.
I wanted a passthrough, where the guest OS would see the NIC and then have exclusive access to it.
When I tried doing that, KVM complained that it could not do this with PCI devices.

Hmm, that had me stumped a bit.
And then I learnt about attaching my VM’s NIC directly to a physical interface.
What I needed to do was use Macvtap as my network source in the VM’s settings.
It then prompted me for my NIC’s device name, which I gave it, courtesy a little spelunking through ifconfig’s output.

And then once I began my Ubuntu VM install, it detected my NIC and then let me assign an ip that I use on my local network.
For all intents and purposes the VM now appears as just another machine on my network ala my Pis.

To my mind, that’d be one less layer of networking I’d need to deal with. ↩︎

Kubernetes on Janusworx

Kubernetes & CNCF Awards

Kubernetes, Note to Self: Need Load Balancer Installed on Bare Metal

K3s: Move Data Folder

Problem #

Solution #

The Plan for My New Hetzner VM

Update: 2025-12-11 #

Preamble #

Goals #

What I want #

2025-11-06 09:35 #

2025-11-06 16:00 #

On How the Kubernetes v1.29 Logo Came About

Partners in Crime aka Spitballing aka Brainstorming #

Why Mandala? #

Putting It All Together #

P.S. The Easter Egg #

Learning Kubernetes, Setup Notes

Make KVM Use a NIC in Bridge Mode

Problem
#

Solution
#

Update: 2025-12-11
#

Preamble
#

Goals
#

What I want
#

2025-11-06 09:35
#

2025-11-06 16:00
#

Partners in Crime aka Spitballing aka Brainstorming
#

Why Mandala?
#

Putting It All Together
#

P.S. The Easter Egg
#