5.4-STABLE corrupting fs on shutdown -p now?
c0ldbyte at myrealbox.com
Mon May 2 05:31:35 PDT 2005
-----BEGIN PGP SIGNED MESSAGE-----
On Mon, 2 May 2005, Steven Hartland wrote:
> Kris Kennaway wrote::
>>> Surely thats what the buffer sync is doing. Watching the console both
>>> these actions appear to be doing identical things but the end results
>>> are quite different :( buffers are always under 5 usually 2 and its not
>>> like there's been a great deal of activity. The last test I did was just a
>>> boot, login, shutdown ( result was / not clean but all the others ok ).
>> Modern IDE drives sometimes lie about having finished writing out the
>> data, when in fact it's still only in cache. You should be able to
>> disable this behaviour to test (try the BIOS).
> I have a "HDD Post Write Buffer" in the BIOS according to the MB manual
> although the disks aren't connected to the MB controllers they are on RAID 5
> set off a PCI-X card. Will give it a go tomorrow ( drag the monitor
> back to the rack ). Will this have any performance penalties? I could
> just avoid using shutdown -p but it seems a bit dodgy what its doing.
> If this machine where miles away without remote console it would have
> caused major amounts of pain.
Why would you use shutdown -p on a remote machine anyway. I could see
possibly a shutdown -r but -p come on.
( When in doubt, use brute force. -- Ken Thompson 1998 )
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (FreeBSD)
Comment: Fingerprint = D1DC 0AA4 1C4E EAD4 24EB 7E77 B261 50BA F7DF 979F
-----END PGP SIGNATURE-----
More information about the freebsd-hackers