1 |
On 06-Aug-12 10:50, Kraus Philipp wrote: |
2 |
|
3 |
> I'm using a software raid with mdadm (mirror). A few days ago the |
4 |
> mdadm removes during the boot process one disk and sets the raid |
5 |
> inactive. The disk contains no errors (smartctl) and nothing is |
6 |
> reported in the logs. I reassemble the disk and activate the raid |
7 |
> again. Reboot the system and my disk are in sync and everything works |
8 |
> fine. |
9 |
> |
10 |
> I'm a little bit confused about it. Does anybody has got an idea, why |
11 |
> the array lost a disk? |
12 |
|
13 |
It may happen quite frequently with common (non-raid) drives |
14 |
which do not have TLER implemented. |
15 |
|
16 |
Modern drives have some kind of internal error-recovery |
17 |
procedure which covers remapping of bad/weak sectors. This |
18 |
might take quite long time (a few seconds, or even tens |
19 |
of seconds) and as a result raid-controller (or raid-software) |
20 |
marks drive as failed because it does not respond in |
21 |
given time. |
22 |
|
23 |
Solution is to use "raid-edition" drives, with TLER |
24 |
(time limited error recovery) implemented. On some drives |
25 |
this can be activated by using some utility (i.e. WDTLER.EXE). |
26 |
You can read more about it on the net, i.e.: |
27 |
http://en.wikipedia.org/wiki/Time-Limited_Error_Recovery |
28 |
http://en.wikipedia.org/wiki/RAID#Problems_with_RAID |
29 |
|
30 |
Jarry |
31 |
-- |
32 |
_______________________________________________________________ |
33 |
This mailbox accepts e-mails only from selected mailing-lists! |
34 |
Everything else is considered to be spam and therefore deleted. |