(Li|U)n[iu]x admins collected and own WTFs
-
@Random832 said:
@Nik_Doof said:
Heres quite a recent one:
A friend of mine moved /lib/glibc* to get his compiled program to use a glibc local to the program, while we were in the process of fixing that minor issue, a few months before I did a cleanup of installed deb packages, and removed dhcpcd. The problem is, the box is coloed and gets it's IP via dhcp, doh.
ouch, how'd you fix it (the glibc thing)? (i know one way - there's a statically-linked copy of "ln" that only creates hardlinks on most systems, you can run it if you have an shell open.)
Smart sysadmins keep a statically linked copy of busybox installed, in case one of the critical core components gets eaten by killer weasels or buggy disk controllers.
-
@Random832 said:
ouch, how'd you fix it (the glibc thing)? (i know one way - there's a statically-linked copy of "ln" that only creates hardlinks on most systems, you can run it if you have an shell open.)
netboot, thankfully the colo offers a network rescue boot (hence the dhcp ip assignment). Just a matter of mounting the local drive, move files etc. :)
-
@joe_bruin said:
I found out the hard way that on Solaris, the killall command does not take any arguments.
My new cowoker discovered it the hard way on a production server a few weeks ago too :) Are you sure your name is Joe?
Personally,
deleted partition table with important datas. I knew the partition layout (first reiserfs, then swap, then ...) but not the exact sizes. Challenge was to manually set partition table, checkfs, on error, try othe boundaries, etc. When you have a 512 byte granularity and know the size at about 500Mb, how much tries does it requires? Had to manually apply efficient search algorithms
seen it:
While using a priviledged user account on ancient company, sudoed root added a new account for future access, changed his password, to discover he had change the root password of a server he does not own and should not be using anymore. Not knowing the original root password, the computer was left for a few month with NO root password :)
-
@Nik_Doof said:
Heres quite a recent one:
A friend of mine moved /lib/glibc* to get his compiled program to use a glibc local to the program, while we were in the process of fixing that minor issue, a few months before I did a cleanup of installed deb packages, and removed dhcpcd. The problem is, the box is coloed and gets it's IP via dhcp, doh.
I did that when manually moving from Linux 2.0.x to Linux 2.2.x. I was on libc5 at the time, and had to upgrade it, so I compiled it by hand. The instructions said to be very careful about installing the new one, but I did a 'mv libc.so.5 oldlibc.so' and then everything broke. Fortunately, I had a boot cd and this was my local machine.
The first time I played with fork() I discovered what a fork() bomb is. Quite amazing, actually.
I learned Linux from a book on general Unix admin, which had a section explaining the 'killall does one thing on Linux, and something completely different on everything else' noted in it, so I missed out on that fun. :)
Finally, I only recently had the 'rm <somethingIwantedtokeep>' problem. I was upgrading to Slackware 12.0, and I was cleaning up the .new files under /etc. I meant to type 'rm dh<tab>.new', but instead hit <enter> instead of '.'. Fortunately, I have my /etc directory under source control so 'svn revert <file>' is an instant undelete. (Source controlled /etc is awesome all around. It's a bit of an experiment, but it's great when you get it working.)
-
My best one was years ago on Redhat. I was trying to adjust the delay on something at login. I honestly don't remember what it was now. However, what I wound up doing was accidentally setting my account so I had just 1 SECOND to type my password in. So you'd type your name in, password prompt would appear, and a second later you'd be staring at a login prompt again.
After many MANY attempts to type my password in quick enough, I eventually managed it on about the 50th attempt.
-
I admit it, I'm a Linux noob. I've never used any Linux other than Puppy Linux (awful!).
So a few days ago, I decided to try Debian on a VM.
I have an awful Internet connection.. so I downloaded the 40MB business card distro.
Started installing, waited ages for it to download+install the base system.
Then, I realised I had selected no tasks! So, all I had was a shell.
Time to reinstall, this time I read the manual and decided to just use tasks Standard + KDE.
So.. I installed until it got to the base system point. I had to go.. so I decided to leave it installing. When I return, it's at the exact same point: but I had chosen a mirror nearer to me, which didn't work. So much for that. So I chose another and finally started installing. Over 3 hours to download and install everything.
Then, I enter. So far so good.
The day later, after installing some stuff (packages are good!) I shut it down.
Later I start it again.. but KDE refuses to log in. Hours of digging around on the Internet.. until I realise I had ran out of disk space. So I enlarged the VMDK, enlarged the partition.. and finally I got it to work perfectly. That's where I am now.
-
Treeki> If you're using the plain text editor you have to put in your own br or p tags.
Also why did you try DSL? You could also mail order certain distros (there's actually a site dedicated to that iirc).
-
Hi, It's actually http://www.ubuntu.com, notice the .com
-
Grr, so used to distros and other linux related sites using .org. ubuntulinux.org redirects you to ubuntu.com though.
