I just bought a new 6TB WD Red Plus drive to replace a couple older
drives (one of which had generated an uncorrectable error email from
smartd). I decided that I'd stick it in an external USB-3 SATA drive
"dock" thingy from Thermaltake and do some testing before installing
it.
Using smartctl, I ran a "short" self test and a "conveyance" test, and
both passed.
Then I started a "long" (extended) self test. According to the drive,
that should take 640 minutes (a bit under 11 hours). It almost
immediately reported test running with 90% remaining. A couple
hours later, it still said 90% remaining. I stopped the test, ran
another short test. Messed around with some other smartctl commands,
and started another "long" test. Again, it immediately said 90%
remaining. After 20 hours, it still said 90% remaining.
Tests getting stuck like that seems to be pretty common.
Some time spent Googling found me two suggestions:
* The test is stalled because the disk is busy.
* The test is stalled because the disk is idle.
Apparently, they're both valid.
The self-test runs in the background when the drive is not busy, so if
the drive is heavily loaded, self tests can take a lot longer. But,
if the drive is _completely_ idle, it might spin-down and go to sleep
(which pauses the test). This is reportedly more likely to happen when
the drive is attached via a USB-SATA adapter [I don't know if that's
true].
Sure enough, if I waited 5-10 minutes and asked for the test status, I
would hear the drive spin-up when queried.
So I ran a "watch" command to query the drive every 10 seconds:
watch -n10 smartctl -d sat -c /dev/sdc
That seems to keep the drive spinning without doing any actual R/W
ops, and within an hour the status changed to 80% remaining. About 6
hours after that, it's now 20% remaining.
So the moral to the story is: when running an extended self test on a
drive, you don't want it to be busy, but you also don't want it so
idle that it spins down goes to sleep.
Maybe everybody else already knew that, but it took me an hour or two
to figure it out...
drives (one of which had generated an uncorrectable error email from
smartd). I decided that I'd stick it in an external USB-3 SATA drive
"dock" thingy from Thermaltake and do some testing before installing
it.
Using smartctl, I ran a "short" self test and a "conveyance" test, and
both passed.
Then I started a "long" (extended) self test. According to the drive,
that should take 640 minutes (a bit under 11 hours). It almost
immediately reported test running with 90% remaining. A couple
hours later, it still said 90% remaining. I stopped the test, ran
another short test. Messed around with some other smartctl commands,
and started another "long" test. Again, it immediately said 90%
remaining. After 20 hours, it still said 90% remaining.
Tests getting stuck like that seems to be pretty common.
Some time spent Googling found me two suggestions:
* The test is stalled because the disk is busy.
* The test is stalled because the disk is idle.
Apparently, they're both valid.
The self-test runs in the background when the drive is not busy, so if
the drive is heavily loaded, self tests can take a lot longer. But,
if the drive is _completely_ idle, it might spin-down and go to sleep
(which pauses the test). This is reportedly more likely to happen when
the drive is attached via a USB-SATA adapter [I don't know if that's
true].
Sure enough, if I waited 5-10 minutes and asked for the test status, I
would hear the drive spin-up when queried.
So I ran a "watch" command to query the drive every 10 seconds:
watch -n10 smartctl -d sat -c /dev/sdc
That seems to keep the drive spinning without doing any actual R/W
ops, and within an hour the status changed to 80% remaining. About 6
hours after that, it's now 20% remaining.
So the moral to the story is: when running an extended self test on a
drive, you don't want it to be busy, but you also don't want it so
idle that it spins down goes to sleep.
Maybe everybody else already knew that, but it took me an hour or two
to figure it out...