Is there a point of scrubbing after a resilver?

[RoboKaren]

Most of the material on zfs notes that resilvering is a form of scrubbing -- in at least that the data and parity data is re-read, recalculated, and written out to the resilvered drive.

But is the data that is written to the resilvered drive re-read to check that it was written correctly?

Is there any need to scrub after a resilver to make sure that the new drive is read/writing properly, or is this automatically done as part of the resilver?

 

[SweetAndLow]

no need to scrub after a resilver. But you should be doing your standard maintenance scrubs.

 

[jgreco]

You'd be better off forcing the new drive to do a SMART long test from the CLI. Scrub/resilvers only touch allocated areas of the disk.

 

[RoboKaren]

Thanks, will run:

sudo smartctl -t long /dev/ada0

after the resilver then.

 

[danb35]

...and the SMART self-tests, like the scrubs, should be scheduled to happen on a regular basis.

 

[RoboKaren]

...and the SMART self-tests, like the scrubs, should be scheduled to happen on a regular basis.

That's all good in principal but is there an official way of doing a long SMART self-test? What I did was:

sudo zpool offline mypool gptid/xxxxxxx-xxxx-xx-xxxxx-xxxxxxxxxxx
sudo smartctl -t long /dev/ada0

and then waited the 572 minutes for it to long self-test a 4TB volume. But I can do this because "mypool" is a zfs-2 and taking one drive offline is feasible. But what do other folks do? Do you cron the long self-tests? Or just dedicate one weekend a month to taking the whole pool offline?

 

[danb35]

You don't need to offline the disk. Just schedule the tests through the FreeNAS web GUI. It's under Tasks.

 

[SweetAndLow]

That's all good in principal but is there an official way of doing a long SMART self-test? What I did was:

sudo zpool offline mypool gptid/xxxxxxx-xxxx-xx-xxxxx-xxxxxxxxxxx
sudo smartctl -t long /dev/ada0

and then waited the 572 minutes for it to long self-test a 4TB volume. But I can do this because "mypool" is a zfs-2 and taking one drive offline is feasible. But what do other folks do? Do you cron the long self-tests? Or just dedicate one weekend a month to taking the whole pool offline?

you're are making that way to complicated just use the gui.

http://doc.freenas.org/9.3/freenas_tasks.html#s-m-a-r-t-tests

 

[RoboKaren]

Hahaha, when did that (the GUI item) appear?

 

[danb35]

At least by 9.0, but it hasn't always been in the same place. The move to "Tasks" happened with 9.3.

 

[SweetAndLow]

Hahaha, when did that (the GUI item) appear? :)

You should never have to do anything outside of the gui, if you are then you are doing it wrong.

 

[jgreco]

You should never have to do anything outside of the gui, if you are then you are doing it wrong.

I guess I do lots of things wrong.

 

[danb35]

You should never have to do anything outside of the gui, if you are then you are doing it wrong.

That's probably an overstatement; there are plenty of things that can't be done from the GUI (like fixing upsd.conf, for example)--but if it can be done from the GUI, and SMART test scheduling certainly can, it should be.

 

[RoboKaren]

Are we sure that resilver does a read-verify? I'm looking at the iostats for my resilver operation and it doesn't show any significant reads from that device (line 18: lots of writes, no reads):

Code:

/
/
/ Zpool iostat
/
                                           capacity     operations    bandwidth
pool                                    alloc   free   read  write   read  write
--------------------------------------  -----  -----  -----  -----  -----  -----
freenas-boot                            5.28G  9.22G      6      0  14.1K  1.42K
  mirror                                5.28G  9.22G      6      0  14.1K  1.42K
    gptid/xxxxxxxxxxxxx-8bc8-11e4-xxxx-xxxxxxxxxxxxx      -      -      4      0  8.35K  1.44K
    gptid/xxxxxxxxxxxxx-3ea1-11e5-xxxx-xxxxxxxxxxxxx      -      -      2      0  5.97K  1.44K
--------------------------------------  -----  -----  -----  -----  -----  -----
kiyoshi                                 3.33T   765G    586    100  55.7M   595K
  raidz2                                3.33T   765G    586    100  55.7M   595K
    replacing                               -      -    827     85  22.3M   368K
      gptid/xxxxxxxxxxxxx-dae0-11e3-xxxx-xxxxxxxxxxxxx      -      -    447     16  14.0M   157K
      gptid/xxxxxxxxxxxxx-a3ff-11e5-xxxx-xxxxxxxxxxxxx      -      -      0    685     13  22.6M
    gptid/xxxxxxxxxxxxx-a1e6-11e5-xxxx-xxxxxxxxxxxxx      -      -    152     10  14.1M   154K
    gptid/xxxxxxxxxxxxx-4028-11e5-xxxx-xxxxxxxxxxxxx      -      -    424     16  14.0M   155K
    gptid/xxxxxxxxxxxxx-a39c-11e5-xxxx-xxxxxxxxxxxxx      -      -    110    231  3.35M  10.8M
    gptid/xxxxxxxxxxxxx-7375-11e4-xxxx-xxxxxxxxxxxxx      -      -    152     10  14.1M   154K
    gptid/xxxxxxxxxxxxx-68a1-11e5-xxxx-xxxxxxxxxxxxx      -      -    155     11  14.1M   156K
--------------------------------------  -----  -----  -----  -----  -----  -----

 

[Arwen]

In someways a scrub after a re-silver can be helpful. (If not already scheduled...)

Specifically;

1. When using mirror with more than 1 source disk, (aka 3 way or more)
2. When using RAID-Z2 or RAID-Z3
3. To confirm the newly re-silvered data

For the first 2, I would assume a re-silver only reads one redundant source, and if
good, write the re-silvered data. So a re-silver is not a scrub.