Notifications
Clear all
Topic starter
I have a ThunderBay 4 in RAID 5 config of 24TB. It is being time machine backed up once daily to a ThunderBay 6 in RAID 6 config of 30TB.
I have Time Machine auto backups set off, and use Time Machine Scheduler to make 1 backup every night at 1AM. So the ThunderBay 6 sits idle for the better part of the day. It's probably idle 23.5 hours of a day.
Would it be better for the life of the drives in the ThunderBay 6 to spin down after backup, and sit non-spinning for the 23.5 hours, then spin up to complete the backup, and the unmount and spin down again?
If the consensus is no, just keep them spinning, then what if I said "I only backup once a week, at 1AM on Sundays." Now is it better to spin the box down, maybe even take it offline and kill power?
Thanks for your advice.
Posted : 28/11/2019 11:51 pm
There are multiple studies on this and none show any difference in low duty cycle, high duty cycle, 24 hour power on, vs low power on hours.
So yes you save electricity. But it does not reduce wear and tear on the drives for some reason.
Posted : 29/11/2019 1:28 am
Topic starter
Thank you for the reply. I also have read a bit on the topic, and most people agree on two things:
1) spindle bearing wear is the major player in HD wear
2) spin up is very stressful for the bearing
But the issue I have when I read the opinions, is that most people are figuring that the HD would be spinning up 5-6 times a day, perhaps only down for an hour or two.
I'm thinking that maybe since I'm literally talking about a spin up once a day, it might be better to spin down. Hmmm.
Ancillary question: if the drive is spun down (but powered) does that count towards drive hours count in SMART?
Posted : 29/11/2019 11:09 am
A spun down drive that is powered will count towards its power on hours, yes.
The major studies on drive longevity and failure (for example the google study of 100,000 disks) do not show any relationship between light/heavy duty and failures. Power up/Power down cycles also do not seem to have any statistical relationship to failure.
You do quote "opinions", but I have not seen any published data that demonstrates power cycles, or spin cycles are related to failures. If any exists, please publish it, I would love to see it.
SoftRAID added a feature to collect disk information, and if we find any correlations, we will certainly publish that data. So far, the only statistical correlation predicting future failure are reallocated sector events. Power cycles/spin cycles do not seem to have a meaningful correlation. This could change with more data.
Posted : 30/11/2019 3:00 am
Topic starter
Thank you for your detailed reply.
Posted : 30/11/2019 11:01 am
