1 |
"P.V.Anthony" <pvantony@×××××××××××.sg> posted |
2 |
463017AD.7050401@×××××××××××.sg, excerpted below, on Thu, 26 Apr 2007 |
3 |
11:08:29 +0800: |
4 |
|
5 |
> Hi, |
6 |
> |
7 |
> Just got the Marvell 6145 chip to work using the patch from, |
8 |
> http://marc.info/?l=linux-ide&m=116986924301674&w=2 |
9 |
> |
10 |
> Now I need to test to see if it works under stress conditions for my |
11 |
> hard disks. |
12 |
> |
13 |
> Please share some links to applications that can do some stress test for |
14 |
> hard disk. Please note that the application should be command line as |
15 |
> there is no GUI like KDE or Gnome. |
16 |
|
17 |
After having a hard drive go out on me a couple years ago, the second one |
18 |
in two years (that one overheated, after the AC went out here in Phoenix, |
19 |
so the room temp could well have been 60 C or higher, and no telling what |
20 |
the drive was), I decided to buy four SATA drives and RAID them together |
21 |
using the kernel RAID (md/mdp). However, before I actually started |
22 |
installing anything on them, I ran badblocks on them, beginning to end, |
23 |
data destructive test (thus the before anything critical is on them is |
24 |
vitally important, tho badblocks does have a couple less intensive modes |
25 |
that aren't data destructive), multiple patterns, repeat-write one |
26 |
pattern one end of the disk to the other, read it back and verify, repeat- |
27 |
write a second pattern, read and verify... IIRC there were four patterns |
28 |
in all, so eight times over the entire 300 gig, four writes, four reads, |
29 |
1.2 TB per drive, four drives, so 4.8 TB written and read. |
30 |
|
31 |
Talking that sort of data transfer volume, it obviously took /days/, tho |
32 |
luckily I was able to test the four of them in parallel without adding to |
33 |
the time much, since the bottleneck was the disk read/write speed, not |
34 |
the bus or CPU speed. Of course, that's 100% sequential access, so it |
35 |
didn't stress the seek mechanisms much, but it's also 100% duty cycle, / |
36 |
days/ at a time (two and a half days, IIRC), so it certainly well tested |
37 |
the data path, that is, the chipset, the low level drivers, the i/o path |
38 |
electronics on the drive itself, the read/write heads and disk platters. |
39 |
All four disks passed without incident, so I felt reasonably secure in |
40 |
using them for mixed RAID. FWIW, RAID-6 for the main system, so I can |
41 |
lose two drives without losing data. |
42 |
|
43 |
badblocks is part of e2fsprogs, so there's a good chance you have it |
44 |
already merged. =8^) The badblocks manpage tells you what you need to |
45 |
know to run the tests, both destructive and not. |
46 |
|
47 |
-- |
48 |
Duncan - List replies preferred. No HTML msgs. |
49 |
"Every nonfree program has a lord, a master -- |
50 |
and if you use the program, he is your master." Richard Stallman |
51 |
|
52 |
-- |
53 |
gentoo-amd64@g.o mailing list |