New 'zpool status' under 11.1

[danb35]

Just upgraded my server to 11.1 the other day, and its first scrub started this morning. zpool status is showing something new to me:

Code:

  pool: tank
 state: ONLINE
  scan: scrub in progress since Mon Dec 25 00:02:03 2017
   5.02T scanned at 212M/s, 2.77T issued at 117M/s, 42.4T total
   0 repaired, 6.53% done, 4 days 02:40:44 to go

The amount scanned I get, but "2.77T issued" is new to me. But that's what seems to be determining the percent complete. What is it?

 

[joeschmuck]

I noticed the same thing yesterday and I have not found anything that references this except I noted that zfsonlinux also has this text in zpool status during a scrub but no description of this yet.

Is the 5.02T scanned the amount of data scanned and the 2.77T issued the amount of data transferred as data. What I might be saying here is the 5.02T is all data including all parity whilst the 2.77T is just the data minus the parity. Honestly I have no clue, just a stab in the dark. I would like to know the true meaning of course.

 

[Ericloewe]

he amount scanned I get, but "2.77T issued" is new to me. But that's what seems to be determining the percent complete. What is it?

It's the sequential scrub/resilver. Since it's done in two steps (scanning and issuing).

First, a buffer is filled up (it defaults to 10% of main memory, IIRC) with reads that are to be issued. As it fills up, the larger chunks of contiguous reads are issued in as few passes as possible, to minimize the IOPS needed. Periodically, it also completely flushes the buffer and updates the legacy scrub status variable, so that it can be resumed later if the system shuts down (even on older versions of ZFS).

 

[joeschmuck]

Is there a place that describes this? Don't get me wrong, I'm please to see your answer, I just wanted to know if there is something else that explains it maybe better. Im not following completely. Maybe it's too much Egg Nog?

 

[Ericloewe]

Video: https://www.youtube.com/watch?v=SZFwv8BdBj4&feature=youtu.be\
Slides: https://drive.google.com/file/d/0B5hUzsxe4cdmVU91cml1N0pKYTQ/view

 

[Ericloewe]

Note: ZFS default seems to be 5% of RAM.

 

[joeschmuck]

Video: https://www.youtube.com/watch?v=SZFwv8BdBj4&feature=youtu.be\
Slides: https://drive.google.com/file/d/0B5hUzsxe4cdmVU91cml1N0pKYTQ/view

That was actually an interesting presentation. Thanks for the links!

 

[Scharbag]

Video: https://www.youtube.com/watch?v=SZFwv8BdBj4&feature=youtu.be\
Slides: https://drive.google.com/file/d/0B5hUzsxe4cdmVU91cml1N0pKYTQ/view

That was a great presentation. Thanks for sharing.

 

[Xelas]

If I'm reading this presentation correctly, does that mean that when you resilver it is actually re-ordering the blocks into logical order on the new drive? Thus doing what is effectively a defrag? So the implication is that you can remove/wipe/add/resilver the drives sequentially and defrag a pool if you really need to?

 

[Ericloewe]

If I'm reading this presentation correctly, does that mean that when you resilver it is actually re-ordering the blocks into logical order on the new drive? Thus doing what is effectively a defrag?

No. Reads to the exact same blocks are reordered to allow for more sequential disk usage. Nothing is rewritten.