A Sysadmin’s Lot Is Not A Happy One

Green Room Window

Man, if anyone ever asks you why so many sysadmins have such a hate on for Windows (and honestly, who doesn’t want to have an answer ready for a question like that?) one of the reasons you can give them is that having to patch a whole dependency-tree full of various stuff is just a huge pain in the ass. Why won’t updates just restart the relevant services on a windows box yet? Because coming from a Unix background, man, having to restart a machine just to update a service is clearly ridiculous. But here we are! (And updating clustered machines? Damn.) Whoever you are, world, this is why I missed your party, and why I left work at three thirty in the morning to fight the dregs of halloween for a cab home.

I’m really looking forward to having centralized logging at work, so that maybe we’ll be able to tell what’s actually going on, and which of these errors are real or not; Asp.NET stuff, for example, spooges all over my logs for no apparent reason at all. Dear Programmers: Error messages should mean something, you know? When some service doesn’t start, and that’s a totally expected condition then it shouldn’t spit out errors like it’s the end of the world. And conversely! Who wants to open up their logs at three in the morning and find something that says “ASP.NET# failed to futzpop bangwarble antwerp AF03 zarqzarq: additional information in error message”, particularly when the “additional information in error message” is “00 0C”. Thanks, I’ll get right on that!

Microsoft, I’m looking at you: this is why there are so many people who hate and fear your server software, and nobody really loves it; when things go south, there’s nothing whatsoever in there to clue you in as to what’s actually gone wrong much less how to fix it. If you ever sit back in your Aerons wondering idly why your support forums are full of bizzare inquiries and quasi-superstitious suggestions (I tried these eight unrelated things, and then lit two candles! One of them worked and I don’t know why, but try them all including the candles!) this is the reason. Error messages are incredibly important public-facing information, and they’re less intelligible now than they were on NT 3.5.

I was going to direct that at Oracle too but, man, screw those guys.


  1. Posted November 1, 2008 at 5:22 pm | Permalink

    I’m glad to see you included Oracle at the very bottom there, because after installing an Oracle cluster, there’s one thing I know for sure: Oracle error messages are really obscure and useless. Even worse is when you pay thousands a year to get access to the support system, and they repeat the “text” of the error message at you.

    Moving on, I agree with the whole patching = restart dilemma. If I’m in the middle of an overnight render, I don’t want my computer to magically restart itself. The price I pay for security and not using Linux, although Slackware is looking increasingly attractive.

  2. Posted November 1, 2008 at 7:23 pm | Permalink

    Yeah well…Mac OS X ain’t much better on this front. It gets a pass because no one seriously uses it for anything more than a nice desktop. My personal favorite Windows error message? The service failed to start because the service started properly. I love that one. Really.

  3. mhoye
    Posted November 1, 2008 at 10:38 pm | Permalink

    Yeah, the Oracle stuff is just completely opaque. And the worst part is, they make so much of their money off courses and certifications that there’s no reason for them to make it less opaque, only more.

  4. Paul Collins
    Posted November 3, 2008 at 3:24 am | Permalink

    On Windows you cannot update a file that’s “in use”, DLLs and EXEs often falling into this category. So Windows updates build a list of files to be replaced on the next reboot, which they then strongly recommend to the user happen very soon.

    Keywords: MoveFileEx, MOVEFILE_DELAY_UNTIL_REBOOT, PendingFileRenameOperations.

  5. mhoye
    Posted November 3, 2008 at 12:15 pm | Permalink

    That’s an excuse, not a reason. You don’t actually reboot to finish up your pending file rename operations; the OS knows what files are open by what processes! Those processes can be shutdown, the pending copies completed and restarted without actually restarting the entire OS from scratch.

    Honestly, I can see that being a reasonable claim when they were pushing to get windows update out the door a decade ago, but not now.

  6. Paul Collins
    Posted November 3, 2008 at 6:38 pm | Permalink

    I used to think Windows knew which files were being used by which processes, but after a few botched WebSphere updates (all together now, ha ha WebSphere) I’m no longer so sure.

    I completely agree that if Windows didn’t suck then it would not suck.