I cam across a similar discussion on the btrfs mailing list a couple
of months ago. The argument went that since PostgreSQL does copy on
write, was it feasible to switch off copy on write for the directories
holding the database? The answer in that case was no, because btrfs
apparently doesn't do checksumming for files with COW disabled.
Aargh! (and I'll see if I can dig the link out of my browser history
later on today).
On 17 Jun 2015, at 13:08, InterNetX - Juergen Gotteswinter wrote:
yep, for sure the same applies to compression and probably other
settings. but i whouldnt touch checksumming, never.
Am 17.06.2015 um 12:55 schrieb Perttu:
What about compression and checksumming? Those also seem redundant
inside the guest and could be turned off for better performance.
-Perttu
On 17 Jun 2015 at 13:49:30, InterNetX - Juergen Gotteswinter
([email protected] <mailto:[email protected]>) wrote:
With with default Settings you will end up in Caching everything
twice
with ZFS inside the Guest.
Either use some regular, legacy FS or change primary/secondarycache
to
Metadata only
Am 17.06.2015 um 12:20 schrieb David Finster:
Hi Perttu
We don’t run any Linux inside KVMs, but we do run a fair amount
of
Windows. In general, your right in that running ZFS inside a KVM
isn’t
going to get you any benefit and would probably be detrimental. It
is
worth noting that all writes from inside a KVM are considered
synchronous anyway as far as the ‘zones’ pool is concerned. Any
writes
that your guest does will be immediately committed to some disk
(hence
the importance of an SLOG).
ZFS snapshots are consistent when created, but I guess there is the
potential for some application inside the VM to have writes
outstanding
when a snapshot is taken (through queued IO?). If that were the
case,
then you would need some mechanism of notifying the application
that
it’s about to be snapped to be completely safe.
That being said, anything that uses a proper transaction log should
be
fine. I’ve done snapshots of SQL and Exchange servers and
migrations/test emergency restores have been fine. Doing a snapshot
of a
live VM and then bringing it up on another host is similar to it
experiencing a power failure, which is what the transaction logs
are
designed to handle.
As Ian mentioned, you might also be better off giving LX zones a go
-
they are much friendlier to the host and no additional file systems
are
involved.
- Dave
On 17 Jun 2015, at 8:00 PM, Perttu <[email protected]
<mailto:[email protected]>> wrote:
Hello SmartOS community, long time follower, first time poster
here.
What do you guys think are best practices regarding file systems
in
KVM Linux guests?
Is it advisable to use ZFS inside guests? Wouldn’t it use double
the
memory for the same data?
Or would using something like XFS or ext4 be better and perhaps
tuning
them to be more synchronous if crash resiliency is wanted?
My main concern is data integrity if I snapshot and send the guest
regularly to another host for disaster recovery. Performance is
also
important but I do have a slog device (S3700).
Cheers,
Perttu
*smartos-discuss* | Archives
<https://www.listbox.com/member/archive/184463/=now>
<https://www.listbox.com/member/archive/rss/184463/25738179-216c4b5f>
| Modify <https://www.listbox.com/member/?&;> Your Subscription
[Powered by Listbox] <http://www.listbox.com/>
*smartos-discuss* | Archives
<https://www.listbox.com/member/archive/184463/=now>
<https://www.listbox.com/member/archive/rss/184463/26910370-9cc4a721>
|
Modify
<https://www.listbox.com/member/?&;>
Your Subscription [Powered by Listbox] <http://www.listbox.com>
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed:
https://www.listbox.com/member/archive/rss/184463/27159464-c7cd6100
Modify Your Subscription: https://www.listbox.com/member/?&;
Powered by Listbox: http://www.listbox.com
*smartos-discuss* | Archives
<https://www.listbox.com/member/archive/184463/=now>
<https://www.listbox.com/member/archive/rss/184463/26910370-9cc4a721>
|
Modify
<https://www.listbox.com/member/?&;>
Your Subscription [Powered by Listbox] <http://www.listbox.com>
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed:
https://www.listbox.com/member/archive/rss/184463/27157529-01fe90e8
Modify Your Subscription:
https://www.listbox.com/member/?&;
Powered by Listbox: http://www.listbox.com
--
Paul Sture
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription:
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com
