Tech

That sucked...

Free time, a nagging error, and almost catastrophic results. How not to spend a day off.

Geoff Anderson

05 Dec 2025 — 2 min read

Alas, sometimes, a "quick" fix really turns into a shitty day.

This is going to be a whine about technology, and someone who knows just enough to be dangerous[1] who about screwed the pooch.

A few months back, the update of the underlying system that this publication runs on, Ghost, had a major update. That update necessitated the upgrading of one of the components, node.js.[2] so I followed the instructions and upgraded from 18.x to 22.18.0 (a long term stable version).

Alas, being a bit of a spaz, I left some remnants of the old version, and every time I went to do a routine update, I got a warning that there was an old version.

Since I am taking a week off of work, I figured this would be an easy thing for me to fix. All the recommendations I researched recommended that I move to the nvm (nodes version manager) system to make it easier and intuitive to manage.

So, I stripped out the old stuff, added the nvm package, and followed the directions to get the version I needed up and running.

And it fucking barfed on me.

Suddenly, I had two non-functioning publications, and a headache that was killing me.

I got the original version of Node.js installed. No dice.

I rebuilt the system. No dice.

I went to the systemd services descriptions. No dice.

I was poking around the log files, and it was not very helpful. I was getting error 203 thrown, and searching the internet didn't help one iota.

I am going to admit that I was a bit chuffed, and about this close to just saying "fuck it" and just pulling it all down. (this close -><-)

What was the problem?

Turns out that nvm maintains the node versions in the user's home directory, and no matter how I set up the permissions, the systemd daemon couldn't execute the runtime file.

The fix was somewhat straightforward, once I figured that out.

I stripped the nvm bits out.
I reinstalled the original nodejs package and defined the version I wanted.
I edited the systemd service files with the new location.
I had to force systemd to reindex (that took some sleuthing in the logfiles)
I had to force a new update to make sure that the underlying ghost system was all copacetic.

Fingers crossed, and it started up. Both properties.

And finally no more node version warnings being thrown.