1 |
Den 19. des. 2015 01:01, skrev Dale: |
2 |
... |
3 |
> It sounds like the heads are doing random |
4 |
> reads/writes and the heads are moving but doing so noisily. Thing is, |
5 |
> there is no drive activity according to gkrellm or iotop. All the |
6 |
> drives should be basically idle. |
7 |
... |
8 |
> Model=ST3000DM001 |
9 |
What kind of file system ? How full ? Raid ? lvm ? What kind of SATA |
10 |
controller? There is lots of stuff about these things I don't know, but |
11 |
I /do/ know all those cause different quirks and telling more specifics |
12 |
would maybe prod someone with experience with your particular type of |
13 |
set-up to chime in. |
14 |
|
15 |
Just as an example, ext4 will delay some of the work on a newly |
16 |
formatted drive. Lots of other system activity can cause background |
17 |
jobs. If this coincides with a bad sector on a drive, you get noise, |
18 |
access latency and system wait. This system wait will not necessarily be |
19 |
counted against any specific process in "top(1)" (depending on how your |
20 |
kernel was compiled) , so you can even get 0%idle and still have no |
21 |
specific process sticking out in "top". |
22 |
|
23 |
Watch your logs and kernel messages for access errors. Difficulties in |
24 |
reading might cause retries without actually timing out. If this is your |
25 |
problem, it will persist until that sector is marked as failing AND |
26 |
SOMETHING IS WRITTEN TO IT. The sector will not be mapped out until |
27 |
something gets written to it. |
28 |
|
29 |
Also, if you are running raid, google around for permutations of |
30 |
"smartctl -T permissive -l scterc,70,70", seeing as this drive |
31 |
<http://www.seagate.com/www-content/datasheets/pdfs/desktop-hdd-8tbDS1770-7-1511UK-en_GB.pdf> |
32 |
is not a server type drive. But, remember, NEVER blindly run something |
33 |
on your system that some random guy sent you by email without |
34 |
understanding what it does. :-D . |