Another case of the vanishing disk
Darren Pilgrim
list_freebsd at bluerosetech.com
Sun Mar 16 18:02:26 UTC 2014
On 3/15/2014 11:04 PM, cruxpot wrote:
> Back in December, it was the power supply. That was a cheap Rosewill
> 300W PSU. The new is a Corsair CX500 (500W). The system basically just
> has an old SCSI card and 4 Green Barracuda 2TB disks and a low end
> pci-e video card and pci-e gigabit NIC. How can the PSU be the problem
> since I replaced it and it's more than adequate?
How are the drives connected to the power supply? Are they all on the
same rails or are they spread across mutliple sets of rails?
Be aware that you may have shot yourself in the foot buying "green"
drives. Drives not designed for use in NAS/RAID usually have firmware
that expects the machine to sleep the disks and be tolerant of delayed
responses. The drives get to be cheaper because the controller has more
"offline" time to fix errors due to higher tolerance parts. In some
cases (like certain WD disks), the drives eventually start dropping off
the port because they're going into an offline error recovery mode and
take too long to respond. On a regular desktop, the OS knows to wait
because the drive was signalled into a sleep mode. That doesn't happen
in a server and you really don't want it to happen in a server.
I'm betting that even if you had each drive on its own +3.3v, +5v and
+12v rails, a line-interactive UPS and a server-grade power supply,
you'll still have dropouts.
More information about the freebsd-questions
mailing list