serving_myself
Dabbler
- Joined
- Jan 2, 2023
- Messages
- 12
Hello there,
I'd like to enable ALPM for my SATA HDDs by changing the
I wasn't able to find a sysctl for this and trying to run
I'd use powertop for that, but the old version of it on Scale (from 2019) has this set to
github.com
Could you please let me know what's the best method to do that?
P.S: I'm aware that this was changed to
However, based on my research so far I see that these were with slower disks (SMR and/or 5400rpm) which in theory could be at play as well. I can also see that for some people who had issues with
This is new build w/o any data yet (only TM backup is sent there so far) so I'm willing to test this in my own env for the sake of low(er) power.
I'd like to enable ALPM for my SATA HDDs by changing the
/sys/class/scsi_host/host*/link_power_management_policy
from max_performance
to med_power_with_dipm
or med_power
(if the former will result in a degraded pool).I wasn't able to find a sysctl for this and trying to run
midclt call system.advanced.update '{ "kernel_extra_options": "link_power_management_policy=med_power_with_dipm" }'
+ reboot wasn't fruitful.I'd use powertop for that, but the old version of it on Scale (from 2019) has this set to
min_power
, and this was changed later to med_power_with_dipm
in this commit:tuningsysfs: use med_power_with_dipm for SATA link power management · fenrus75/powertop@02703a0
Instead of using min_power, which has reports of data corruption, use med_power_with_dipm (Device Interface Power Management) available since 4.15. This has the same behaviour as the Windows IRST ...
Could you please let me know what's the best method to do that?
P.S: I'm aware that this was changed to
max_performance
at some point in Scale because some people's pools became degraded and they had bunch of SATA errors in their logs.However, based on my research so far I see that these were with slower disks (SMR and/or 5400rpm) which in theory could be at play as well. I can also see that for some people who had issues with
med_power_with_dipm
, the med_power
was just as OK as max_performance
.This is new build w/o any data yet (only TM backup is sent there so far) so I'm willing to test this in my own env for the sake of low(er) power.
Last edited: