ATA Drive Issues
freebsd at juiceless.net
Mon Apr 3 18:28:32 UTC 2006
> From: "Wil Hatfield" <freebsd at hyperconx.com>
> I am giving 6.1 a whirl.
> In the first 5 minutes I have already noticed that there are some
> obvious filesystem issues fixed. I ran a tar and compared the
> speed to that on one of my 4.x boxes and low and behold they are
> about the same. THANK GOD!
I had similar issues with a server freezing, and it also turned out to
be WRITE_DMA failures on one of the (SATA) HDs. This was on 6.0-release,
and upgrading to 6-stable helped. Nice to see it fixed your problem,
too. 6.1 seems to be a really stable and performing branch.
When reporting problems for HDs, it's good policy to always state the HD
and motherboard/chipset brand, and to point out if/when one has
checked/renewed the cabling (as you did). Also test results from
smartmontools and HD manufacturer disk tests are good to state.
Even when doing this, the FreeBSD community is peculiarly keen on
blaming everything on failing HW and bad cabling. My experience is,
though, that the first thing to blame (when cabling and HDs have passed
tests), is the HD controller, especially when it's one of the crappy
ones, like Silicon Image 3xxx. These crappy chipsets results in that
FreeBSD doesn't have good support for them, as they are unreliable per
I don't meen to critisize FreeBSD, on the contrary. But it's a bit
annoying to always get responses to all HD related stuff that "your HW
is failing", when everything you've done for the last week is tests
which claim it isn't.
I'm not familiar with the innards of FreeBSD, so could someone explain
why FreeBSD is so picky about cabling? I'm pretty well-read in physics,
so the theory of FreeBSD using less-energetic electrons to pass data in
the cables doesn't seem to hold up (insert smiley here). I mean, if
other operating systems don't do some sort of error checking to exclude
the errors on HD/CD-ROM drives that make FreeBSD hiccup, then what's up?
More information about the freebsd-questions