From: T. Horsnell (tsh@mrc-lmb.cam.ac.uk)
Date: Fri Jan 13 2006 - 06:24:34 EST
Many thanks to all who responded, and I'm glad
to see that Tom Blinn is still looking after us.
The overall consensus seems to be that Seagate
drives are well regarded, and that I should
pay attention to power and cooling.
I pretty much did that before embarking on this
rolling disk replacement procedure - the drives
are housed in BA350 shelves (remember them? the
gunmetal blue jobs) with 7 drives to a shelf,
sharing a 180Watt PSU.
The max power-requirement for these drives
(at startup) is 14Watts of 12V and 3.5Watts
of 5V. i.e. no more that 20Watts, so 7 drives
(140Watts absolute max) should easily be handled
by the 180Watt psu, assuming its working OK.
Temperature is a bit more difficult to monitor
(Oh for something like Linux's smartctl.
Can I do this with scu?). Before I put a
new drive into production, I run my own test
procedure on it which consists of write/readback/compare
random length records of random data to random disk
positions. I do this on the raw device to
avoid the UBC, and these days I also turn
off the on-disk cache. I run this test continuously
for 48 hours or so, and then open the plastic
disk caddy and feel the drive temperature.
They just feel warm. The computer-room is
aircon'd to 18C (I used to run it at 16C but
the heat-exchangers kept icing up and our maintenance
dept got upset).
So I dont think power or temp is an issue.
My fileserver has about 100 drives on it of
varying sizes and antiquities. (I just pensioned
off the last SCSI2 drive left over from
our Alpha 2100 we bought some 10 years ago,
the ones in the biege caddies). The smaller
(and slower) drives seem to be much more
reliable than the more recent bigger faster
ones.
I have 14 x Seagate 180GB 1" profile drives (I've
forgotten the model number) which have been running
for a few years with essentially no problem.
I have 28 of the 140GB Cheetah running for up to
1.5 years and have had 1 fail with a big burst of
unrecoverable read errors, and one with what
I took to be a bearing failure (pulsating whine).
I have 7 of the 300GB disks, one of which was
DOA (couldnt read or write to it), another
read back all zeros, whatever was written to it,
(this too was out-of-the-box), one failed with
tons of uncorrectable read errors, and another
has just done the same.
Maybe they are from the same batch. I dont know
how to tell.
These all fail the Seagate Seatools tests, and
get replaced under warranty, but I use almost all
my disks in JBOD mode, and I'm getting VERY nervous.
SO - has anyone any experience of JetStor products?
(http://www.raid.com) or similar?
Apologies for the diatribe.
Terry.
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:50:27 EDT