Today's article is about journalctl...

[KGIII]

This is a pretty complicated subject and there was no way that I'd ever cover it all. So, I told AI to do it for me - but just the basics were covered. I think this might have even been too much for AI to cover everything.

Demystifying journalctl: A Comprehensive Guide to Linux System Logging • Linux Tips

It was suggested that I write an article about journalctl and it really seemed like a large topic. Here's a journalctl article.

linux-tips.us

Also, one of my plugins hosed my site for like five hours today. It took the whole site down. This is my fault for automating stuff, but the automation is good for security. So, it's a mixed bag. I do check the site often enough so that I could probably disable automated updates, but I don't want to miss a 0-day event and end up with a compromised site.

As I've said before, security is a process. I've picked mine. I know the risks. I'd rather a site outage once a year (that takes maybe 20 minutes to debug and fix) than risk getting hacked and, worse, losing control of the little information I do have about other people. I value your privacy more than my frustration. In theory, I'd use a staging site that was configured just like the one you see. I'd then do all sorts of testing on the staging site. Ah well...

 

[blunix]

<span style="font-family: helvetica, arial, sans-serif;">journalctl --since "2024-04-01 00:00:00" --until "2024-04-15 23:59:59"</span>

html freaked out here

 

[KGIII]

Yup.

It wasn't like that during the preview stage but my site went haywire today. So, that might have been the cause. It's an easy fix. Thanks!

 

[KGIII]

And it should be fixed. Again, thanks.

It looks like I could have missed it during the preview stage? I'm not sure if the glitch caused that, though the timing is similar. So, it could have been the glitch. The timing is close, but I should have seen an error screen and not the preview.

Hmm...

Yeah, I'm going to blame the glitch. A WordPress plugin went haywire, making the site completely unavailable. I had to go in to enable debugging, rename the plugins directory, and figure out which plugin caused the problem.

It turned out to be a superfluous plugin that had 3 releases within just a few hours because it was causing sites to throw an error. It's just a plugin that monitors queries and displays things like load speeds for the various calls to the database or PHP. The initial of those updates was suspiciously close to the time I pressed the 'schedule' button on the article.

Meh...

Thanks. It's fixed.

 

[dos2unix]

I use journalctl -xeu (service.name) frequently.

 

[KGIII]

I use journalctl -xeu (service.name) frequently.

I had to look up the -x. That's interesting. I only use the basics, thus the article. I really can't talk about advanced usage. I thought about having the AI do a longer and advanced article, but the output from my prompts wasn't great and I wouldn't really be able to verify it if it got too advanced.

 

[blunix]

I just wrote a LONG blogpost on systemd timers and the fantasy nonsense chatgpt and ollama came up with was hilarious.

That being said, some features I wrote about I didn't find good documentation for at all so I had to manually try them to see the results.

Anyways, using LLM atm for anything more than rather basic Linux stuff is not really a thing. Last times I broke it was over ceph, nginx and systemd timers.

 

[KGIII]

I just wrote a LONG blogpost on systemd timers and the fantasy nonsense chatgpt and ollama came up with was hilarious.

I saw your long recent article. That took hours to write!

Anyways, using LLM atm for anything more than rather basic Linux stuff is not really a thing.

Yeah, it really didn't do well with some other prompts. In fact, when asked for an advanced version it wrote me an article that didn't have a single line of example code. Not one single code snippet was included...

I tried a variety of prompts and this one returned the closest to what I might have written had I tried it myself. I only wanted to cover the basics - instead of spending 6 hours writing an article.

 

[blunix]

I saw your long recent article. That took hours to write!

3 days xD The annoying part was to actually test all the badly documented config options and then putting examples into the blogpost so that less experienced admins can understand it..
There is a systemd timer option that lets you trigger a timer upon someone manually changing the time on your system.
brah.
who needs that.
I grepped through /etc/systemd and didnt find any default timer in ubuntu that uses that. idk. bit weird. systemd is an init system including a kitchen sink xD

Yeah, it really didn't do well with some other prompts. In fact, when asked for an advanced version it wrote me an article that didn't have a single line of example code. Not one single code snippet was included...

its a bit about the prompt. you have to tell it to "not write about marketing or convince people, only write technical arguments and provide examples by code lines".

Journalctl is a cool thing though.. I have a blogpost about central logs with systemd that also covers some journalctl commands. It actually is much cooler than what we had before with rsyslogd (default, or syslog-ng if you are fancy) using syslog ID's to split up logs into files.
Storage is much more efficient too now.
All in all thats a part of systemd I like.

 

[KGIII]

The annoying part was to actually test all the badly documented config options and then putting examples into the blogpost so that less experienced admins can understand it..

I suspect it'd have been easier to do a long in-depth journalctl article as it appears to be documented properly in the man page.

Alas, I really only use the very basics.