Posted by: Karsten M. Self
on February 10, 2005 05:32 PM
The authoritative answer will be in the whitepapers, linked at the Sourceforge site in my previous post.
The tests are AFAIU not based on any destructive tests to the drives. My understanding is that most of the tests are checking historical registers (the drives log errors / issues), and checking performance to see that it's within spec (spin up, head movement, etc). A long test is good for some interesting sound effects....
If you've got an already dodgy disk, there's probably a possibility the the tests themselves will result in failures, so your first step should probably be to ensure your backups are current and complete.
There is a performance impact, particularly for the long test, and default installations tend to run this in what should be off-hours (your own lousy sleeping habits are not a developer problem.
There's a <tt>WARNINGS.gz</tt> file in my Debian install, which desribes a parade of horribles that may happen. Essentially: you can hang your OS fi calls are made on poorly designed hardware.
My own experience includes a drive which was operating far slower than spec, by an order of 150 times. Running the long test was simply not feasible -- I think it ran well over 12 hours before I terminated it (typical is ~20-40 minutes). But I took that as a strong indication that the disk was, in fact, not 100% dependable.
Learn about Linux
Download Linux
Get Linux help
S.M.A.R.T. safety
Posted by: Karsten M. Self on February 10, 2005 05:32 PMThe authoritative answer will be in the whitepapers, linked at the Sourceforge site in my previous post.
The tests are AFAIU not based on any destructive tests to the drives. My understanding is that most of the tests are checking historical registers (the drives log errors / issues), and checking performance to see that it's within spec (spin up, head movement, etc). A long test is good for some interesting sound effects....
If you've got an already dodgy disk, there's probably a possibility the the tests themselves will result in failures, so your first step should probably be to ensure your backups are current and complete.
There is a performance impact, particularly for the long test, and default installations tend to run this in what should be off-hours (your own lousy sleeping habits are not a developer problem.
There's a <tt>WARNINGS.gz</tt> file in my Debian install, which desribes a parade of horribles that may happen. Essentially: you can hang your OS fi calls are made on poorly designed hardware.
My own experience includes a drive which was operating far slower than spec, by an order of 150 times. Running the long test was simply not feasible -- I think it ran well over 12 hours before I terminated it (typical is ~20-40 minutes). But I took that as a strong indication that the disk was, in fact, not 100% dependable.
#