On 07/09/17 11:40 AM, Thomas C. Bishop wrote:
> I have servers configured w/ HW controlled raid and have had virtually
> NO problems w/ those. Both my backup machines are SW raid... I've had to
> replace multiple drives on the SW configured raid. The drives are either
> SAME MODEL or same Seagate drive family in all cases and one server is
> actually the same SuperMicro model as one of the desktops.
>
> I had attributed this to just a hotter running environment.. the backup
> machines are desktop workstations w/ NVIDIA graphics cards that run
> pretty hot, but I'm rethinking this now.
>
> Any chance SW raid is running the HDs harder/hotter than the HW raid?
> All machines run 24-7-365 so power cycling is not the issue and the
> server room is not necessarily cooler than the office/desktop environment.
>
> Tom
I would argue software raid is going to run your drives harder then a
battery backed raid card.
The cards DRAM buffer will probably shift a large majority of writes to
full stripe writes. Vs. if you do anything with files smaller then
stripe basically EVERYTHING is going to be a read-modify-write on md raid5.
All that said, is it going to be enough of a workload delta to see
lifetime differences? That's going to depend on your workload. I have
quite an old array and my drives seem to not care so... YMMV.
# for drive in sda sdb sdc sdd sde sdf sdg sdh; do smartctl --all
/dev/$drive|grep Power_On_Hours; done
9 Power_On_Hours 0x0032 027 027 000 Old_age Always
- 64114
9 Power_On_Hours 0x0032 035 035 000 Old_age Always
- 57735
## the raid5 ##
9 Power_On_Hours 0x0032 090 090 000 Old_age Always
- 49785
9 Power_On_Hours 0x0032 022 022 000 Old_age Always
- 57543
9 Power_On_Hours 0x0032 084 084 000 Old_age Always
- 80950
9 Power_On_Hours 0x0032 022 022 000 Old_age Always
- 57364
9 Power_On_Hours 0x0032 098 098 000 Old_age Always
- 1078
9 Power_On_Hours 0x0032 098 098 000 Old_age Always
- 1079
--
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
