Time Machine Failure

[...] This post was mentioned on Twitter by Mark Grimes and Samuel's Feed, Abdul Jaleel K.K. Abdul Jaleel K.K said: Time Machine Failure http://bit.ly/hwEi3z [...]

I think you'd have to be nuts or a newbie to rely on Time Machine as your primary backup system as opposed to a more straightforward cloning system. But Time Machine is actually a pretty cool versioned secondary backup system with appropriate hooks into the OS, has real world uses, and I bet it works just fine 95%+ of the time.

Of course, I think we could assemble a quorum to agree that it'd have been nice if Cupertino had continued to develop OS X after 2007, but that's a whole 'nother topic...

FWIW, if I understand things correctly, doing a boot into Single User Mode will force Time Machine to do a "deep traversal" backup the next time you fire up backupd-helper. (And my understanding is that there is no simpler way to force a "deep traversal" backup.)

So, in a sense, the fact that I always do a SUM reboot to run AppleJack to run fsck when I reboot "solves" the problem Snyder outlines.

Of course, my current Uptime is 42 days, so it's a damn good thing I do regular clones in the meantime in addition to Time Machine...

Been a Mac user for many a year and a Unix administrator for years as well. My only backup is Time Machine. I rotate 2 external drives onto the machine (each drive alternates a month as the Time Machine drive and a month in a fireproof safe.)

It is my experience that any backup strategy has a weak point. The weak point of most is the convenience of doing the backup. Time Machine is very convenient and the one time that I had to use it, it worked flawlessly - much better than having no backup.

"The weak point of most is the convenience of doing the backup. Time Machine is very convenient and the one time that I had to use it, it worked flawlessly - much better than having no backup."

This is very true.

However, you are already going waaaaaay beyond what is "convenient". You're rotating 2 drives and dealing with a fireproof safe. How much of an additional inconvenience would it be at that point to add an automated clone routine every time you plug in a new drive to your current Time Machine routine? CCC and SD are both fine pieces of software for accomplishing that task.

Time Machine is certainly better than nothing for newbies and casual users. But for pros, it's best thought of as a cool supplementary versioned backup system with good hooks into the OS. Once you start jumping through a few sane extra hoops to do your backups, adding in cloning seems like a pretty minor additional hoop to jump through.

I think of Time Machine as well executed technology that is also Space Shuttle-ish in its complexity. That bargain brings with it a success rate in excess of 95%, but well short of 100%.

But it's your data and your time, not mine. We all decide on our own tradeoffs.

I don’t understand why they didn’t design Time Machine to do an automatic daily deep traversal as a safety measure. Even Spotlight does this.

"I don’t understand why they didn’t design Time Machine to do an automatic daily deep traversal as a safety measure."

Considerations for resources on older hardware? Considerations for the time someone might want to connect their bus powered backup drive on their laps? Consideration for weak wireless networks?

I can understand the design decision tradeoffs of not doing daily deep traversals by default. In fact, if there were no way to turn that theoretical feature off, I'd prefer Time Machine to not do daily deep traversals, since not all my hardware consists of latest model boxes.

My confusion is that there seems to exist no command line method to force Time Machine to do a deep traversal pass, short of the reboot-SUM method I mention upthread.

"Even Spotlight does this."

Now, this part, I wasn't aware of, and perhaps don't fully understand.

I stay to the surface to the point where I thought Time Machine and Spotlight were working off the same index.

(Although, I suppose the fact that they aren't finally explains to me why rebuilding your Spotlight index doesn't force a Time Machine deep traversal, which has always confused me.)

Spotlight isn’t working from an index; it is the index (and totally separate from Time Machine). Both Spotlight and Time Machine are working from filesystem events. However, Spotlight assumes that it may not be able to keep up with the changes as they happen or that there might be changes that it didn’t get events for, and so it does a full traversal once in a while. Even, I assume, to index Time Capsule backups. So my point is, why is this worth the resources for a content index but not for a backup? If anything, that seems backwards.

Your point makes sense.