Replication question/problem

[Z300M]

My original pool (Pool1, two striped six-disk RAIDZ2 vdevs) suffered a major meltdown (three disk faults in one of the vdevs) when I fired the system up again after a six-week vacation, and many files obviously did go missing in action: they still appeared in the directory, but any attempt to copy them to another drive failed. But I have copied to external drives what I can copy and the GUI shows Pool1 with 9.7TiB used. I don't recall what the "used" report was before the meltdown.

I have created another pool (Pool2, five-disk RAIDZ2), and set up a replication task to Pool2/Replicas with the intention of destroying Pool1, recreating it as an 11-disk RAIDZ3 vdev, and restoring everything from Pool2, BUT Pool2 shows only 2.7TiB used. So should I assume that the indicated 9.7TiB for the space used on Pool1 is bogus and that there is really only 2.7TiB of files there?

 

[Robert Trevellyan]

Are there a lot of snapshots on Pool1?

 

[Z300M]

Are there a lot of snapshots on Pool1?

1268 items, but "Used" for many of them is zero.

 

[Robert Trevellyan]

1268 items, but "Used" for many of them is zero.

That could account for at least some of the difference.

 

[Z300M]

That could account for at least some of the difference.

I ran

# du -ch /mnt/Pool1 | grep total

and got

9.0T total

but that presumably includes the snapshots, yes?

Any command that would tell me the total EXcluding the snapshots?

 

[Z300M]

That could account for at least some of the difference.

I ran

/mnt/Pool1# du -ch --exclude=/.zfs/snapshot/* * | grep total

and still got

9.0T total

then

/.zfs/snapshot# du -ch * | grep total

gave

20G total

 

[Robert Trevellyan]

It looks like the zfs command might be more useful than du. The man page talks about various subcommands that begin with "used" that look relevant.

 

[Z300M]

It looks like the zfs command might be more useful than du. The man page talks about various subcommands that begin with "used" that look relevant.

I realized that I had not checked the "recursive" box. I had executed ls on Pool2/Replicas, seen the subdirectories, and assumed that they were populated -- but I assume that without the "recursive" parameter it would have created the subdirectories but not their contents.

I created a one-time snapshot and did a zfs send | zfs receive operation. All seems to be well.

BTW, I couldn't figure out how to use

zfs list

with the usedbysnapshots parameter, but

zfs list all

includes the usedbysnapshots report.

 

[Robert Trevellyan]

Are you saying the numbers you're seeing now make more sense?

 

[Z300M]

Are you saying the numbers you're seeing now make more sense?

Very definitely.

 

[Robert Trevellyan]

:)