On 07/11/17 08:48, Phil Turmel wrote:
On 11/06/2017 04:38 PM, Eyal Lebedinsky wrote:
On 07/11/17 07:15, Wolfgang Denk wrote:
Dear Eyal,
In message <c12a2a32-2321-1ed7-e1de-ce0e408552e1@xxxxxxxxxxxxxx>
you wrote:
(I had many disk failures/replacements in the 4 years life of
this array).
Reading this makes me wonder if you checked your environment for
other influences. there must be some reason for an exceptional
high number of failures.
I remeber we also had a nighmare of disk errors in the rack in the
2nd floor of our building - which disappeared after moving the
rack into the basement. I can't prove it, but I blame it on
vibrations. We have a heavy traffic train line less than 50 meters
away, and disks (classic, magnetic ones) definitely do not like
vibrations - see [1]. Maybe you have other influences you did not
check for yet?
Interesting Wolfgang,
- This array is at home, a relatively quiet place. - I monitor the
disks temperatures and it is OK. - The machine runs off a UPS which
can be a source of bad power (if the PS does not filter it out). -
The HBA may be somehow bothering the disks?
The disks are under warranty until late next year so there is time to
see if the disks do better with the LSI.
BTW, two of the RMAs were for disks that arrived DOA (as RMAs). I do
not have high regard for the WD blacks. The failures were spread
across the last 4 years (so not infant mortality).
If nothing else, this experience made me comfortable with software
raid, and encouraged me to stick to my backup schedule.
That's a really bad failure rate. But they're WD Blacks, which if I
recall correctly, do not support scterc. Did you deal with your driver
timeouts?
Yes, my rc.local does
# echo 180 >"/sys/block/$disk/device/timeout"
The disks were let go after developing bad sectors (more than once with
increased rate). I did not have a case of a disk kicked out due to lack
of scterc.
I *will* get more suitable disks next time.
If not, those drives probably weren't really dead. Just not
raid-compatible out-of-the-box.
Phil
--
Eyal Lebedinsky (eyal@xxxxxxxxxxxxxx)
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
