Tuesday, December 12, 2006

Lessons from another XP disk crash

Another year, another 3 disk deaths. Laptops are murder on drives. I lost two in my work (Dell) laptop, one from my home (iBook) laptop. Only the most recent Dell death was suspicious, the other drives were about 3 years old and they'd been worked far beyond reason. Laptop drives are not designed to be flogged continuously, they can't get rid of the heat fast enough.

Still, the cost of these dead drives is high; the lost productivity costs on the XP drives probably exceeded the costs of the laptops -- much less the drives. There was no data loss thanks to my backup obsession and some advance warning of each failure, but the time required to restore a complex XP work environment is daunting. The big problem with XP is that critical user data files and metadata are scattered all over the drive and registry, but a 'clone' restore isn't well supported by most automated backup systems. OS X is vastly better of course, but still imperfect -- I'm most interested in what 10.5 will do to speed this process.

A picked up a few more XP lessons from this recent series. Here they are:
  • Funny noises are a common warning of impending disk failure, but in the latest case I was experiencing some inexplicable problems with network connectivity. It was very hard to figure out what was causing these. Turns out bits of the disk were dying, taking out XP system components.
  • The clue that my software unreliability was due to a failing drive were seek errors on my backups. Retrospect Pro didn't makes as big a deal of this as I would have liked, but at least I knew enough to chase down the report. Errors on backup systems always need to be investigated, and a 'seek error' is a mark of doom.
  • The windows Event Viewer (note this web page, reviewed 9/06, is missing about half its content. I think Microsoft has outsourced its knowledge base management to Apple.) was catching the disk errors, and quietly recording them. Did it scream a warning every time this happened? No, why should it? Warnings like that might distract me from Microsoft's Vista marketing effort. Lesson: Use filter settings on the event viewer to show only alerts and warnings and check it once a week. I'll keep an eye out for a utility that generates a real alert for me based on checking the event viewer log, I'm sure one exists for XP. I'm also going to take another look at XP disk error monitoring utilities.
  • Most backup software, when it encounters a disk error, just stops. Of course one would prefer it continue and get as much as it can, then announce the disk error in red letters several inches high ...
  • If you delete an XP user profile, it doesn't go to the trash. It's just gone. Data recovery software works well on this kind of goof, however. If you ever do this, turn off the machine immediately and take the drive out! (Ok, so this was only tangentially related to the drive crash. I was naively/stupidly following the advice of tech support when trying to sort out the network errors that were, in fact, drive failure errors). My backup was a bit out of date (see fail on drive error, above), so we tried data recovery software first (OnTrack) and it worked very well.
  • When copying files from an old drive or backup repository to a new drive Windows copy is way too slow and unreliable. xcopy has a switch to ignore errors (important given those seek errors); it works but doesn't log the errors -- so you don't know what to fetch from the backup. I prefer robocopy.exe (free from Microsoft, see resource kit, below). Here's the command line example: ROBOCOPY D:\WORK E:\ROBOCOPY /V /TEE /S /E /COPY:DAT /B /R:5 /W:2 /NP /LOG:E:\ROBOCOPYLOG_061117B.TXT. Note OnTrack will do offset read/writes to try to get things ROBOCOPY can't, but ROBOCOPY is free. I ran robocopy. exe on my flaky drive before resorting to backup, the bad sectors only knocked out one data file which was easy to restore.
  • When sorting out the new machine, start by downloading and installing XP Power Toys (I always need to run TweakUI to fix the worse annoyances of XP, and install the power tab switcher and command-here, etc) and the Win server 2003 resource kit.

No comments: