There are 2 kinds of people in the world, those who love the command line and those who don’t. I fall into the former category. Because I was a MS-DOS baby (okay, probably 5 or 6 years old), who started my computing journey typing
deltree into that lovely white on black prompt trying to figure out how to run games.
P.S. don’t arm a kid with
deltree and leave them unsupervised.
Some people find it odd that a front-end developer, a CSS-lover no less, likes to play around with servers and sysadmin-related activities. I also enjoy building my own machines, sourcing for parts and all that. Stop categorising people into little boxes, I say.
The Ubuntu upgrade that went south
Anyway, I have a home-made low-powered NAS running under my desk which serves all my media files, and it was running Ubuntu 16.04 since I built it. But now that 18.04 has been out for a bit, I thought I’d do an upgrade last night.
In hindsight, it probably wasn’t the best idea to run
sudo do-release-upgrade at 9.30pm in the evening, but I’m not known for making good decisions. The issue with my server is that I’ve had this long-standing issue with mdadm whereby the package is broken but I cannot purge nor reinstall it.
Setting up mdadm (3.3-2ubuntu7.1) ... dpkg: error processing package mdadm (--configure): subprocess installed post-installation script returned error exit status 20 Errors were encountered while processing: mdadm E: Sub-process /usr/bin/dpkg returned an error code (1)
Stack Overflow tried to help but to no avail, but given that I can still serve files off the server, I’ve left it alone. For more than 2 years. And now it sort of came back around to bite me on the ass, because this issue interrupted the upgrade process.
I wasn’t sure what happened exactly but it seemed like the upgrade went along halfway but when I ran
sudo apt-get update and
sudo apt-get upgrade, after a reboot, there were loads of packages that needed to be updated, so I think the upgrade went through.
The shell problem
Unfortunately, my fish shell went haywire, not sure what happened there exactly, as it appeared there was some issue with my oh-my-fish installation broke as well. I thought removing and reinstalling the whole setup would help.
In hindsight, if I did it properly it probably would have. But I obviously did not. It was late. I wasn’t thinking. But instead of removing the package via
apt, I removed the entire
/usr/bin/fish folder without setting default shell back to bash first.
And then, I logged out. 🤦♀️
I had effectively locked myself out of the server as the system got caught in a login loop and could no longer login via ssh anymore. Fortunately, I could directly access the server and login as root via safe boot.
What I learnt about recovery mode
There was some mild panicking when I couldn’t see my default user’s home directory. I thought the entire user somehow got wiped, but when I tried to recreate the user, it already existed.
cut -d: -f1 /etc/passwd showed that my user was still there, just that somehow
/home was completely empty. I was under the incorrect assumption that every user’s home directory was visible to root, but clearly I was very wrong.
It was only when I tried to login as my default user when I realised that deleting the
/usr/bin/fish was a terrible idea. And the way to fix it was to change the default user’s shell back to
bash by editing the
Unfortunately, I ran into this error:
cannot lock /etc/passwd; try again later.
Okay, to the Googlez once more. This time I came away with the command:
sudo mount -o remount, -rw /dev/sda1
Nothing actually happened when I ran this, which I figured was a good thing. Because when I tried to save my changes in the
/etc/passwd file, it worked.
This post was written more for myself than for anyone else, because I’m pretty sure I’ll do something stupid like this again, and so I’ll have something to refer to when I have to undertake similar recovery efforts.
TL:DR, don’t do sysadmin stuff when you’re half-asleep. Bad things can happen.