[OmniOS-discuss] slow system

[Eric Sproul]

Hi Felix,
This sounds like the characteristics of a dedup table (DDT) that will
not fit in RAM.  The consequences this produces are well-documented
and can be found easily via Google.  Constantin Gonzalez has a pretty
good summary here:
http://constantin.glez.de/blog/2011/07/zfs-dedupe-or-not-dedupe

Your options are basically: wait it out It will eventually grind
through all the DDT de-referencing of the freed blocks, or destroy the
pool and start over.

You've learned the hard way that dedup has a dark side.  :)
Eric


On Fri, Jun 21, 2013 at 9:10 AM, Felix Nielsen <felix.nielsen at gmail.com> wrote:
> Hi All,
>
> I am new to the omnios system, but have been using nexenta for a while.
>
> I have an HP Z800 workstation, 96GB ram, integrated LSI controller with IT
> firmware, 6 x 2TB SAS 7200 disks, 2 x 750GB OS mirror, 1 x SSD ARC 160GB
> Intel, 1 x OZ SSD for ZIL - OmniOS and napp-it to handle the ZFS functions.
>
> I started to replicate some snapshots from my Nexenta - 5-8TB in total, and
> I was playing around with compression, dedup to see what gain I would get.
>
> Then suddenly it took for ever to destroy and create new zfs filesystems,
> and replication was really slow.
>
> So I deleted all my replicated volumes and rebooted the system - nothing
> happened, so I was watching the console and it took 4100 sec to shutdown,
> and then it started file system sync, which I am not sure how long will
> take.
>
> What could be the issue?
>
> Thanks
>
> _______________________________________________
> OmniOS-discuss mailing list
> OmniOS-discuss at lists.omniti.com
> http://lists.omniti.com/mailman/listinfo/omnios-discuss
>