Ready, Aye, Ready

I have a serious love/hate relationship with UNIX and LINUX systems. Where I love them is that when they're up and running, they tend to stay that way. However, when they're not, they're nearly impossible to troubleshoot.

Take our current Solaris 5.9 system for example. We've basically retired it with a move to putting the development and bug tracking environment entirely on Windows PCs. This was an essential step for us on two counts. First, the price/performance of the Intel hardware versus the Sun stuff was getting ridiculous. Second, and this is far more important, we ran out of UNIX systems people. There just aren't any left we could use. Being part of a larger organization meant we're required to use the corporate IT group and they don't hire UNIX people. So we're left with these legacy systems that no one knows how to maintain, not even us.

After the new development environment was brought up (took one day and went smoothly, albeit after months of testing and practice runs,) we chose to leave the old Solaris system running in order to make sure that if something went wrong, we had a fall-back position. With no idea of how to bring the computers back up after shutting them down (cross-mounted directories, harddrives with stiction problems, obscure daemon startup processes) this seemed to be the safer deal.

The downside is that some people apparently missed the memo on the move so three weeks after the migration, I'm still getting support questions on why CVS seems to be losing files (they're committing to the old Solaris system rather than the new repository on the PC) or why changes to PRs seem to be lost (same type of issue, but with the problem reporting server.) As responsibility for all of this lands on my increasingly bowed shoulders, I'd finally had enough and elected to go in and just kill off the processes I knew worked reliably on the new system.

Of course, could I figure out how to do this? Not a chance. One server I can't even log into as it's rejecting all incoming network connections and someone lost the keyboard for the connected VT220 terminal (there's no VGA or PS/2 connection due to its age.) The other is accessible and -- glory be -- the root password works, but none of the configuration files are where I'm familiar with from my time on a Linux distro or BSD via OSX, and apparently most of the executables aren't either. I can't even find a CVS daemon to take my frustration out on.

The temptation to go into the server room after the last round and just pull the plug was almost unbearable; heck, no one's ever going to use these things, so a tidy shutdown of the filesystem is unimportant. It just feels like the wrong solution. Still, I'm not being paid to learn how to maintain Solaris so the poor machines may exit rather less gracefully than they arrived.