[Samba] Can't restore from GPFS snapshots, disk_free error

[Ian CLANCY]

Hi Johnathan,
I've been following your reply's to Eric's GPFS Snapshot question with
interest as i'm currently testing a Samba GPFS CTDB setup .
I have a working test system but am curious as to what "gotchas" there are
to running such a system in production ?.
Can you share the benefit of your experience for the list ?.
Thanks in advance
Rgds
Ian

Ian Clancy
IS Department
Valeo Vision Systems (VVS)
Dunmore Road, Tuam,
County Galway,
Ireland.
Tel:  +353 (0)93 23151
Fax: +353 (0)93 25134
Email: ian.clancy at valeo.com
Web: http://valeovision.com/


On 17 October 2013 22:26, Jonathan Buzzard <jonathan at buzzard.me.uk> wrote:

> On 17/10/13 18:49, Eric Chris Garrison wrote:
>
>> Thank you for the suggestions and advice, Jonathan.  We will try out
>> the snapshot options you mention.
>>
>> How onerous are snapshots on a prod system?
>>
>
> When it goes wrong you are looking at file system where basically
> *nothing* happens till you restart the node that is trying to get a
> quiescent lock. Basically your file system is undergoing a denial of
> service.
>
> The advice from IBM was basically "don't do that" unless the file system
> is quiet. Of course given you are running a Samba server and you don't
> control your end users who could of course be doing anything at anytime,
> and especially in a research environment (they keep funny hours) that is
> not exactly helpful.
>
>
>  I think we're going to do
>> just once a day, in the wee hours when not much is going on. What
>> kind of problems do you see?
>>
>
> How does 07:00 on a Saturday morning grab you? I would have thought that
> it would have been pretty quiet at that point, but guess what...
>
> My plan had been to take snapshots at 07:00, 10:00, 13:00 and 16:00 to
> give some roll back for files that got deleted/scrambled during the day
> which would otherwise not exist in the backup. Unless it has improved
> radically in GPFS 3.5 it is a no go in my experience.
>
>
>  How badly does it hang, for how long?
>>
>
> Indefinitely, that is until the node attempting to get a quiescent lock is
> restarted. As you can imagine that is not funny.
>
>
>  What sorts of things aggravate that?
>>
>>
> Basically too much activity on the file system. Problem is you have
> absolutely no way of knowing in advance how much activity is going on.
>
> It looked lovely on my test system. Ran for months without issue. Put in
> production we had lockup's within a couple of days. When in hung at 07:00
> on a Saturday morning I turned it off.
>
> Note there are lots of gotchas running the GPFS/Samba/CTDB combination,
> and this is just one...
>
>
>
> JAB.
>
> --
> Jonathan A. Buzzard                 Email: jonathan (at) buzzard.me.uk
> Fife, United Kingdom.
> --
> To unsubscribe from this list go to the following URL and read the
> instructions:  https://lists.samba.org/**mailman/options/samba<https://lists.samba.org/mailman/options/samba>
>

This e-mail message is intended only for the use of the intended recipient(s).
The information contained therein may be confidential or privileged,
and its disclosure or reproduction is strictly prohibited.
If you are not the intended recipient, please return it immediately to its sender 
at the above address and destroy it.