I have a couple of SATA drives connected via a USB dock that I use only for backups. Normally they are powered down and only come on at night during the backup run. However they also power on any time I reboot the system and have to be powered down again manually. I'd prefer not to power them on at all until they're needed. Is there some magic that would inhibit the boot process from powering up these devices? Note that the /etc/fstab entry already has 'noauto' but that just prevents them from being mounted.

On Tue, 1 Dec 2020 14:52:21 +0100 Roberto Ragusa wrote: > How do you power them down? I don't know about the original poster, but my "noauto" disks timeout and power down all by themselves, then even when they aren't mounted at all, every time I reboot, the reboot hangs while it spins up the disks for some reason. I constantly wish I could make it stop doing that.

On Tue, Dec 1, 2020 at 6:59 AM Tom Horsley <horsley1953(a)gmail.com&gt; wrote: > > On Tue, 1 Dec 2020 14:52:21 +0100 > Roberto Ragusa wrote: > > > How do you power them down? > > I don't know about the original poster, but my "noauto" disks > timeout and power down all by themselves, then even when they aren't > mounted at all, every time I reboot, the reboot hangs while it > spins up the disks for some reason. I constantly wish I could make > it stop doing that. Enumeration. I'm not sure if it's udev scanning everything, or the kernel or both. And also maybe libblkid, to know what everything is. The only way it can know what it is, it needs to be spun up to read the partition map and all the signatures on each partition.

On 12/1/20 8:42 AM, Patrick O'Callaghan wrote: > On Tue, 2020-12-01 at 09:01 -0700, Chris Murphy wrote: >> On Tue, Dec 1, 2020 at 6:59 AM Tom Horsley <horsley1953(a)gmail.com&gt; wrote: >>> >>> On Tue, 1 Dec 2020 14:52:21 +0100 >>> Roberto Ragusa wrote: >>> >>>> How do you power them down? >>> >>> I don't know about the original poster, but my "noauto" disks >>> timeout and power down all by themselves, then even when they aren't >>> mounted at all, every time I reboot, the reboot hangs while it >>> spins up the disks for some reason. I constantly wish I could make >>> it stop doing that. >> >> Enumeration. I'm not sure if it's udev scanning everything, or the >> kernel or both. And also maybe libblkid, to know what everything is. >> The only way it can know what it is, it needs to be spun up to read >> the partition map and all the signatures on each partition. > > Obviously it needs to spin up if it needs to know these things. My > point is that it doesn't need to know them until I tell it. But it can't know that it shouldn't until it knows what's there. :-) The kernel scanning the USB bus will trigger the dock to check the drives which will spin them up. Even if the dock didn't, the kernel would still check the drives, so they would still be started.

On Tue, 1 Dec 2020 13:33:22 -0700 Chris Murphy wrote: > Yep. For sure the kernel is what's responsible for reading the > partition table. And that alone would cause a spin up. e.g. But why would it need to scan when it is shutting down the system (which is my case :-).

I swear it didn't use to do that, then a few releases ago (probably several releases ago) it starting hanging on shutdown till the disks spin up.

On 12/1/20 8:42 AM, Patrick O'Callaghan wrote: > On Tue, 2020-12-01 at 09:01 -0700, Chris Murphy wrote: > > On Tue, Dec 1, 2020 at 6:59 AM Tom Horsley <horsley1953(a)gmail.com&gt; wrote: > > > > > > On Tue, 1 Dec 2020 14:52:21 +0100 > > > Roberto Ragusa wrote: > > > > > > > How do you power them down? > > > > > > I don't know about the original poster, but my "noauto" disks > > > timeout and power down all by themselves, then even when they aren't > > > mounted at all, every time I reboot, the reboot hangs while it > > > spins up the disks for some reason. I constantly wish I could make > > > it stop doing that. > > > > Enumeration. I'm not sure if it's udev scanning everything, or the > > kernel or both. And also maybe libblkid, to know what everything is. > > The only way it can know what it is, it needs to be spun up to read > > the partition map and all the signatures on each partition. > > Obviously it needs to spin up if it needs to know these things. My > point is that it doesn't need to know them until I tell it. But it can't know that it shouldn't until it knows what's there. :-) The kernel scanning the USB bus will trigger the dock to check the drives which will spin them up. Even if the dock didn't, the kernel would still check the drives, so they would still be started.

On Tue, Dec 1, 2020 at 9:42 AM Patrick O'Callaghan <pocallaghan(a)gmail.com&gt; wrote: > > On Tue, 2020-12-01 at 09:01 -0700, Chris Murphy wrote: > > On Tue, Dec 1, 2020 at 6:59 AM Tom Horsley <horsley1953(a)gmail.com&gt; wrote: > > > > > > On Tue, 1 Dec 2020 14:52:21 +0100 > > > Roberto Ragusa wrote: > > > > > > > How do you power them down? > > > > > > I don't know about the original poster, but my "noauto" disks > > > timeout and power down all by themselves, then even when they aren't > > > mounted at all, every time I reboot, the reboot hangs while it > > > spins up the disks for some reason. I constantly wish I could make > > > it stop doing that. > > > > Enumeration. I'm not sure if it's udev scanning everything, or the > > kernel or both. And also maybe libblkid, to know what everything is. > > The only way it can know what it is, it needs to be spun up to read > > the partition map and all the signatures on each partition. > > Obviously it needs to spin up if it needs to know these things. My > point is that it doesn't need to know them until I tell it. Yeah that's not how it works. $ dmesg | grep sda [ 2.149354] sd 0:0:0:0: [sda] 3907029168 512-byte logical blocks: (2.00 TB/1.82 TiB) [ 2.149397] sd 0:0:0:0: [sda] 4096-byte physical blocks [ 2.149514] sd 0:0:0:0: [sda] Write Protect is off [ 2.149545] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00 [ 2.149722] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA [ 2.235611] sda: sda1 [ 2.236610] sd 0:0:0:0: [sda] Attached SCSI disk $ The second to last line, where it discovers the sda1 partition, is needed in order to create all the symlinks for all the various parts of that drive. Not just the dev node, but also the /dev/disk/by- things. If you don't want that to happen, you'd need a udev rule to inhibit that. And then when it goes looking for a particular fs UUID in fstab to mount, of course it doesn't exist, so the mount will fail. It doesn't know it needs to scan everything NOW. There's no correlation between fs UUID in fstab to a particular device until it's scanned, which happens by udev during early startup.

On 12/1/20 2:15 PM, Patrick O'Callaghan wrote: > I have a couple of SATA drives connected via a USB dock that I use only > for backups. Normally they are powered down and only come on at night > during the backup run. However they also power on any time I reboot the > system and have to be powered down again manually. I'd prefer not to > power them on at all until they're needed. > > Is there some magic that would inhibit the boot process from powering > up these devices? Note that the /etc/fstab entry already has 'noauto' > but that just prevents them from being mounted. How do you power them down?

On Tue, Dec 1, 2020 at 9:39 AM Patrick O'Callaghan <pocallaghan(a)gmail.com&gt; wrote: > > On Tue, 2020-12-01 at 14:52 +0100, Roberto Ragusa wrote: > > On 12/1/20 2:15 PM, Patrick O'Callaghan wrote: > > > I have a couple of SATA drives connected via a USB dock that I use only > > > for backups. Normally they are powered down and only come on at night > > > during the backup run. However they also power on any time I reboot the > > > system and have to be powered down again manually. I'd prefer not to > > > power them on at all until they're needed. > > > > > > Is there some magic that would inhibit the boot process from powering > > > up these devices? Note that the /etc/fstab entry already has 'noauto' > > > but that just prevents them from being mounted. > > > > How do you power them down? > > Extract from my script ($SLOT is sd[de]): > >         smartctl --smart=off /dev/$SLOT1 -q errorsonly >> $LOG 2>&1 >         smartctl --smart=off /dev/$SLOT2 -q errorsonly >> $LOG 2>&1 >         echo 1 > /sys/block/$SLOT1/device/delete # Can't use udisksctl because it removes the >         echo 1 > /sys/block/$SLOT2/device/delete # bus, which can't be powered on again > I've got a Seagate laptop drive inside an Intel NUC that is only storage (not system boot or root). And I use a udev rule. $ ls -l /etc/udev/ hwdb.bin hwdb.d/ rules.d/ udev.conf $ ls -l /etc/udev/rules.d/ total 12 -rw-r--r--. 1 root root 139 Oct 7 17:47 60-block-scheduler.rules -rw-r--r--. 1 root root 179 Oct 7 17:36 69-hdparm.rules -rw-r--r--. 1 root root 628 Oct 29 06:40 70-persistent-ipoib.rules $ cat /etc/udev/rules.d/69-hdparm.rules ACTION=="add", SUBSYSTEM=="block", \   KERNEL=="sd*[!0-9]", \   ENV{ID_SERIAL_SHORT}=="WDZ47F0A", \   RUN+="/usr/sbin/hdparm -B 100 -S 252 /dev/disk/by-id/wwn-0x5000c500a93cae8a" $