[rear-devel] New contributors

Schlomo Schapiro schlomo at schapiro.org
Wed Jun 19 13:38:53 CEST 2013


Hello Florent and Yoann and Welcome!

Thank you for your introduction, this helps us to better understand what
you are doing or proposing.

On 19 June 2013 12:48, Yoann Laissus <Yoann.Laissus at bull.net> wrote:

>  Hello all,
>
>  First, let's talk a bit about us and our work. Florent Rochette and me,
> Yoann Laissus, are currently in an internship at Bull (a French-owned
> computer company, bull.com). Currently, a fork of mkCDrec is used to
> prevent disaster for RHEL servers.
>

Is this about servers used internally or do you talk about shipping servers
to customers with a factory installation of RHEL where you add a "stock
image" recovery DVD​​ based on ReaR? In the latter case you might be
interested in contributing the "Backup to DVD" code :-)

Due to the deprecation of mkCDrec, it was decided that Rear will be used in
> a near future. We are responsible of the switch between those two
> solutions. So, we need to make sure everything is correctly supported on
> Bull servers.
>

Do you have a requirements document that you could share? A public
discussion about that might help both you and the project.
​​

>  In the coming months, we will probably send you regular patches to add
> new features or fix bugs we could encounter.
>

Cool!​​ Is this just a project or will you be maintaining ReaR for Bull on
a permanent base?

We have access to servers with many different configurations (eg: Fiber
> channel with multipath, RAID cards, SAN, ...).
>  So we will be able to test Rear with various test cases.
>

What do you think about test automation? One of the big challanges that
ReaR faces at the moment is that almost all testing is done manually. If we
could start to automate that then we could probably roll regular releases
as well.
​​

>  Our first interrogation was about disk matching at the restore time.
> Currently (according to the layout/prepare/default/25_compare_disks.sh
> file), only the disk size is checked. What's your opinion about another
> check like serial number of disk for example (unique identifiers).
>

Do you have numbers on how often a DR happens on the same machine vs. how
often DR happens on another machine (or after swapping disks for
replacements)? IMHO most DR happens onto different disks which for sure
have different serial numbers but probably similar (or larger) disk sizes.

Hence I believe that matching on serial numbers is useless in real life and
would be only useful for test recoveries on the same machine/VM. BTW, if
you test only recoveries on the same machine/VM then you will probably miss
all kind of potential problems. My test setup is a VMware VM as source
system and a VirtualBox VM as recovery target. In that other VM everything
is different...

If you have a use case where ReaR mixes up disks it might be interesting to
persue other means of disk identification:

   - physical connection type (SATA, SAS, IDE, FC ...)
   - shelf location (2nd enclosure - 4th disk) if available

That is all based on the assumption that for physical machines the recovery
is likely to happen on another "copy" of the same machine with the same
parts installed into the same location.

Kind Regards,
Schlomo

PS: Are you related to http://www.bullopensource.org/?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://pikachu.3ti.be/pipermail/rear-devel/attachments/20130619/dd25fe3d/attachment.html>


More information about the rear-devel mailing list