Software, technology, sysadmin war stories, and more. Feed
Friday, January 11, 2013

Thank you for not needlessly reinstalling customer boxes

I love the little nuggets of folklore which sometimes get passed around tech support centers. Some of them have no connection to reality, but they still manage to bounce around a fair bit.

Not long after I started doing my web hosting support thing, I had a customer who had a motherboard replacement scheduled for their machine. I'm don't remember if it had blown a capacitor, was "stabbed" or was just an upgrade, but it was going to happen. They were just another Red Hat Linux box as far as I was concerned.

The maintenance window started and the data center guys did their thing. Then I got a phone call. Apparently they wanted to know "if it would be okay if they just mounted the old drive as secondary". In other words, they were talking about reinstalling the whole box and forcing the customer to migrate their data across... just because of a motherboard swap!

I told them that no, Linux boxes do not work that way, and we do not reinstall them just because we changed a little bit of hardware. The worst thing that could happen is that we might need to adjust the boot loader and/or kernel if there is some specific magic which happened: SMP to UP (back when it was an either-or proposition), different disk geometry due to a BIOS change (yay LILO), or whatever.

I guess this one person had learned somewhere that "new mobo = rekick" and had fully intended to carry it out no matter what. I actually managed to catch a snippet from that ticket for the ages:

Left voicemail with XXXXX. This type of upgrade will require a motherboard change. It would probably be best to do a rekick with new board and mount old drive in server for 2 weeks.

At least this particular tech asked first, and was stopped before anything bad could happen to the customer's machine. It was placed back online with its new motherboard and life went on.

I guess this is the sort of thing which happens for people who have never done this sort of work before. I imagine some of them would be amazed at how long a Linux install can last if you know what you're doing. Here's an example:

$ ls -ld /lost+found/
drwxr-xr-x 2 root root 16384 May 14  2000 /lost+found/

Yep, my workstation box has a 12 year history. Oh, sure, it's been migrated from IDE disks to SCSI and then to SATA over the years, and it moved from being a P133 to a massive beastly thing with several other steps along the way. I even changed it from being plain old x86 to x86_64 at one point just to get the "extra register" benefits for GNU Radio signal processing. Obviously, this would not be possible if I had to reinstall every time a significant bit of hardware changed on the box.

At no point did I just throw it all out and start over like this guy would have done to our customer. That's just too much work.