[kwlug-disc] Was cronjob / btrfs scrup [Was: Re: What is all this about systemd?]

[Chris Irwin]

On Thu, Oct 30, 2014 at 12:54 PM, B.S. <bs27975 at yahoo.ca> wrote:

> I suppose there'd be a performance hit with btrfs checksummed over
> snapshots, but presumably also less complexity, let alone more timely alert
> of any hardware issue. (?)
>

You don't want to store your VMs on BTRFS. They actually specifically say
not to, as you'll cause significant fragmentation. Currently their defrag
method can cause problems with snapshots, I believe.

You can disable COW for your VM directory, which will solve fragmentation.
However, it will also negate the benefits of BTRFS in the first place
(checksums, snapshots, etc). I've still got my "legacy" LVM disk for VM
storage because I can just snapshot the whole filesystem, and either cp,
rsync, or dump the contents.

To meet my goal of as few extra filesystems as possible, I'm trying to
convert my VMs to have NFS root filesystems, which can be exported from my
NAS, and thus benefit from the BTRFS there. I haven't had success yet,
though.


> In poking about, it seemed that such a mirror becomes self-healing. A spot
> goes bad, checksum fails, things get recopied. Interesting.
>

That's why I'm using BTRFS -- checksums and recovery. It's also why I run a
weekly scrub before my snapshot and backup.

ZFS is another option, possibly more stable. However, I wanted to stay
in-tree, especially since my server doesn't have a display connector, so
I'll be blind-booting USB media if I ever have a fail to boot situation.
(or looking for a spare graphics adapter, I guess).

-- 
Chris Irwin
<chris at chrisirwin.ca>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://kwlug.org/pipermail/kwlug-disc_kwlug.org/attachments/20141030/a2356cd7/attachment.htm>