bin

add initial post on the systemdization rss

created pr with 76.1 on 2025-08-23T19:31:20Z · by bcbb7755

cmds

checkout latest patchset:

ssh pr.pico.sh print 76 | git am -3

checkout any patchset in a patch request:

ssh pr.pico.sh print 76.[rev] | git am -3

add changes to patch request:

git format-patch main --stdout | ssh pr.pico.sh pr add 76

Patchset `76.1` on 2025-08-23T19:31:20Z · commit 2e44e9f

→ add initial post on the systemdization

Benjamin Pollack 2025-08-23T19:27:59Z

add initial post on the systemdization

Benjamin Pollack 2025-08-23T19:27:59Z

Semantic diff summary

0 added, 0 modified, 0 signature changed, 0 removed across 0 analyzed files (2 files skipped: unsupported file type)

+2 -0 .codebook.toml #

......@@ -23,11 +23,13 @@ words = [
    "mesos",
    "ndash",
    "ngrok",
    "quadlets",
    "riak",
    "résumé",
    "skunkworks",
    "sourcehut",
    "spreedly",
    "systemd",
    "tkww",
    "xdebug",
]

+54 -0 post/rebuilding-the-lan.md #

......@@ -0,0 +1,54 @@
---
title: "That time I decided to redo the whole homelab"
date: 2025-08-25T09:10:11
draft: true
tags: [
  "tech",
  "devops",
]
---

This is a story about victory and defeat. It's a story about a do-it-yourself
stalwart who insists on understanding every little thing about his computer
deciding to lean into Claude, and promptly really regretting that decision.[^1]
It's about the simplicity of `systemd`, the complexity of Kubernetes, and the
miscibility of those concepts when you get into Quadlets. And finally, it's a
story about me getting more into devops than I have in so many years I had to
bust out a completely new tag on my blog.

But more than any of that, it's a story about just diving in and making
mistakes.

I'm very much not burned out with coding. For the past nearly a year, I've been
back heads-first into the bytes, and I've been loving it. It's not all been
roses---I've dealt with the AI onslaught same as anyone else, except that I also
went about three years without writing any code other than small projects for
fun, so that was great---but my day-to-day has been way better.

What I was burned out on (until this past week) was my homelab. Somewhere in the
middle of the pandemic, when the world was falling apart instead of merely the
United States, I bought a Synology NAS, and filled it to what for me was "the
brim" at about 6 TB. And that much was fine, as far as it went, but I didn't buy
any server to go with it; instead, I just slavishly chucked anything I might be
interested in at

At about this point, I gave up, and simply created a `.volume`, figuring I'd
sort things out later. Which I did, gradually: first, I verified that I hadn't
actually hosed any of the relevant permissions on the actual account. Then, I
verified that setting `PUID` and `PGID` to `0` would do what I wanted in terms
of actual local permissions. Finally, I verified that things did not, in fact,
work.

My next clue came at 16:15, when the clock stopped, and the following clue an
hour later at 16:15 when I discovered the actual meaning of a trailing `:z` on a
volume sharing description! See, with Docker, at some point, I'd begun just
slavishly slapping `:z` on volumes. Monkey see, monkey do, after all. But when I
was converting everything to Quadlets, I dropped it, because "that must be a
Docker thing." But, no, turns out it's an _SELinux_ thing. Specifically, for
reasons I'll have to learn about later, adding `:z` tells SELinux somehow that I
meant to do the thing I'm very obviously deliberately doing. Yay!

[^1]: For those from Lobsters or other similarly minded sites, please don't tag
    this as `vibecoding`, because that's a minor point in this piece, and also,
    while I did not necessarily regret that decision, I did both rue and lament
    it.

...	...		@@ -23,11 +23,13 @@ words = [
23	23		"mesos",
24	24		"ndash",
25	25		"ngrok",
	26	+	"quadlets",
26	27		"riak",
27	28		"résumé",
28	29		"skunkworks",
29	30		"sourcehut",
30	31		"spreedly",
	32	+	"systemd",
31	33		"tkww",
32	34		"xdebug",
33	35		]

...	...		@@ -0,0 +1,54 @@
	1	+	---
	2	+	title: "That time I decided to redo the whole homelab"
	3	+	date: 2025-08-25T09:10:11
	4	+	draft: true
	5	+	tags: [
	6	+	"tech",
	7	+	"devops",
	8	+	]
	9	+	---
	10	+
	11	+	This is a story about victory and defeat. It's a story about a do-it-yourself
	12	+	stalwart who insists on understanding every little thing about his computer
	13	+	deciding to lean into Claude, and promptly really regretting that decision.[^1]
	14	+	It's about the simplicity of `systemd`, the complexity of Kubernetes, and the
	15	+	miscibility of those concepts when you get into Quadlets. And finally, it's a
	16	+	story about me getting more into devops than I have in so many years I had to
	17	+	bust out a completely new tag on my blog.
	18	+
	19	+	But more than any of that, it's a story about just diving in and making
	20	+	mistakes.
	21	+
	22	+	I'm very much not burned out with coding. For the past nearly a year, I've been
	23	+	back heads-first into the bytes, and I've been loving it. It's not all been
	24	+	roses---I've dealt with the AI onslaught same as anyone else, except that I also
	25	+	went about three years without writing any code other than small projects for
	26	+	fun, so that was great---but my day-to-day has been way better.
	27	+
	28	+	What I was burned out on (until this past week) was my homelab. Somewhere in the
	29	+	middle of the pandemic, when the world was falling apart instead of merely the
	30	+	United States, I bought a Synology NAS, and filled it to what for me was "the
	31	+	brim" at about 6 TB. And that much was fine, as far as it went, but I didn't buy
	32	+	any server to go with it; instead, I just slavishly chucked anything I might be
	33	+	interested in at
	34	+
	35	+	At about this point, I gave up, and simply created a `.volume`, figuring I'd
	36	+	sort things out later. Which I did, gradually: first, I verified that I hadn't
	37	+	actually hosed any of the relevant permissions on the actual account. Then, I
	38	+	verified that setting `PUID` and `PGID` to `0` would do what I wanted in terms
	39	+	of actual local permissions. Finally, I verified that things did not, in fact,
	40	+	work.
	41	+
	42	+	My next clue came at 16:15, when the clock stopped, and the following clue an
	43	+	hour later at 16:15 when I discovered the actual meaning of a trailing `:z` on a
	44	+	volume sharing description! See, with Docker, at some point, I'd begun just
	45	+	slavishly slapping `:z` on volumes. Monkey see, monkey do, after all. But when I
	46	+	was converting everything to Quadlets, I dropped it, because "that must be a
	47	+	Docker thing." But, no, turns out it's an _SELinux_ thing. Specifically, for
	48	+	reasons I'll have to learn about later, adding `:z` tells SELinux somehow that I
	49	+	meant to do the thing I'm very obviously deliberately doing. Yay!
	50	+
	51	+	[^1]: For those from Lobsters or other similarly minded sites, please don't tag
	52	+	this as `vibecoding`, because that's a minor point in this piece, and also,
	53	+	while I did not necessarily regret that decision, I did both rue and lament
	54	+	it.

bin

Patchset 76.1 on 2025-08-23T19:31:20Z · commit 2e44e9f

Patchset `76.1` on 2025-08-23T19:31:20Z · commit 2e44e9f