The Brunch Table

Few things are more nerve-wracking than watching an unproven disk utility slowly have its way with your precious data.

Let me back up a bit. A couple weeks ago, after a rash of system crashes, my iBook started taking excessively long (like, 20 minutes) to start up. I booted up into text mode (by holding command-S during startup) and ran fsck, the built-in drive repair utility. (This is what’s running behind the scenes when your Mac takes a long time to boot.) It reported some volume structure problems. It would say something like “you have 203126 whatsits but it should be 203125–fixing it”, but then later on, it said “no, this 203125 whatsits should be 203126–let me take care of that”. Great. My operating system is spending 20 minutes chasing its own tail every time I start up. It seems that there are some situations in which the system can’t heal itself. I wonder: what does Steve Jobs do when his file system starts tearing itself to shreds? Does he just have a stack of fresh 17″ powerbooks in his closet above the row of identical black turtlenecks and blue jeans?

A bit of googling turned up recommendations for a program called Drive 10. Incredibly, it appears to be the only third-party OS X drive-repair program that actually runs in OS X. It also claimed the ability to rebuild volume structures wholesale, which sounded like exactly what I needed to do. I bought myself a copy, and it finally arrived yesterday.

I ran it on Justina’s fresh young iBook, and it came back with a clean bill of health. So far, so good. Then I turned it on my battle-scarred machine, telling it to test everything it could. 95% of the line-item tests in the program were simple, perfunctory affairs that my drive cleared in moments. However, when the program gave me its report, the last section said “Volume Structure: FAILED (-60)”. (What the hell is a “-60″? Note to fellow programmers: either display your information in a form that your users are likely to understand, or don’t display it at all. Seekrit error codes are so 1985.) However, it did give me a Repair button, which I employed with some trepidation. The window also offered the helpful advice that if the repair didn’t work, I should back up my data, reformat the drive, and try again. Lovely.

Now, I was weaned on white rice and MS-DOS Norton Utilities, and I liked watching the wise and powerful Peter Norton move around the little blocks of data on my drive. No such luck here–Drive 10 gives you an attractive image of a wrench sitting on a blueprint-looking thing, and a long blue progress bar, and a couple inscrutable messages of the form “Step 302,342 of 2,880,245″ and “Task 3 of 6″. Fine. With the program cranking away, I left to go eat dinner.

When I came back later, I was dismayed to find that where the “steps” had once been ticking away 150 or so at a time, they were now trickling past, one by one. Great–my drive utility apparently believes in Xeno’s paradox. When I went to bed, the program was still at it. On “Task 5 of 6″, it was now at “Step 2,810,000 of 2,880,245″. Averaging one “step” per second, with 70,000 left–that’s 19 hours and change! Criminey. If this thing ever finishes, after some 24 hours of work, I’d better have God’s own volume structure! My drive had better hum in sweet harmony with the music of the spheres!

At this point, it’s 9:13 a.m., and it still has some 40,000 “steps” left to go. Tune in next time, same bat-time/channel, etc.

Update: It finished counting through its steps at around 5:00 p.m., and then the progress bar turned into a progress-less “barber pole” indicator. It’s still going strong at 9:00 p.m.

Update: Day 3, 7:31 a.m. It’s still on “Task 5 of 6″, but the barber pole has turned back into a proper progress bar, which is now at “Step 2,975,769 of 4,281,281″. The CD that I booted the program from is spinning like mad.

Update: 10:49 a.m. “Step 2,975,853 of 4,281,281″. CD still cranking away.

Update: 1:31 p.m. “Step 2,975,914 of 4,281,281″. Incidentally, I wrote to Micromat tech support, and here’s what they have to say about the situation:

Joe,

Two days much longer than usual to run the standard test–repairs. If there are volume problems on the drive be sure anything important is immediately backed up if possible to be safe. Then proceed with repairs.

A -60 error is an invalid leaf node error. We hope that more descriptive error messages will be included in future releases. Although this may be caused by simple corruption in the directory, it might also be caused by a bad block on the drive. You might want to first do a surface scan test to check for this. A bad block could cause problems when performing a repair.

When Drive 10 does a volume repair it attempts to construct completely new directories in RAM. If it can do this it will provide some feedback in the form of a Technical Comparison about the proposed changes. At that point you can either cancel the rebuild (and nothing will have changed on the drive) or accept it. It is safe to stop the rebuild in the middle since the new directories are being constructed in RAM–the disk is not being altered at that point.

Volumes with OS X installed on them include many tens to hundreds of thousands of files. The volume structures can get extremely complex and rebuilding the directories may take a long time (several hours may be required). As long as the computer has not crashed and you can still move the cursor, then the routines should still be working. In that case we suggest that you let Drive 10 continue. Near the end of the rebuild, as the new directories are being constructed in RAM, you will see the spinning “beach ball” or “barber pole” and it may appear that nothing is happening. The program is doing sorts in RAM at that point. If you let Drive 10 continue we expect it will finish.

A typical time for a rebuild on an average volume with OS X on it is one to three hours. However, in the case of a volume with a very large number or files and severe volume damage we have seen cases where it took several days (and in fact was successful). However, this amount of time is not usual.

Sincerely, MicroMat Tech SupportIncidentally, I did do the surface test first (I’m paranoid like that).

Update: 5/3, 11:56 a.m. When I got up this morning, the CD was still spinning away, and the count had gone up by perhaps 1-2,000 since yesterday morning. I left to have breakfast and do some shopping, and I was ready to just stop the process when I got home. At the rate it was going, it looked like it was going to take weeks, and at that point I just might as well reformat the drive and start from a backup.

However, when I got home, the machine appeared to have restarted itself, and seemed to be in a relatively normal state. This didn’t really square with what the Micromat guy had told me to expect–it never displayed a list of proposed changes for me to sign off on. However, after 3 days of waiting, I was prepared to take what I could get. I installed the Quicktime 6.2 update that was waiting for me, and then I rebooted and ran fsck. This time I got a different bunch of error messages–so Drive 10 must’ve done something. Right now I’m on my second run-through of fsck, and it seems like the error messages are dwindling, so I guess that’s progress.