Gmirror/graid or hardware raid?

[Charles Swiger]

On Jul 8, 2015, at 12:49 PM, Mario Lobo <lobo at bsd.com.br> wrote:
> Hi;
> 
> I'm building a server that is going to host VMs for a small e-commerce
> site. It's mainly going to be used for testing new solutions/improving the
> current site (on the cloud at this moment) but in the medium to long run,
> the idea is to hold the production site in it.
> 
> Since I don't have any experience in running VMs for e-commerce, I don't
> know for sure if my plan is correct or even feasible.

OK.  If it matters, I've been doing e-commerce sites since ~1993, which more
than likely means I have well-entrenched habits about things like HTML validation
rather than in "social media integration" or "cloud-related buzzwords".  :-)

> I want to use FBSD 10 as the host OS and the VMs will run under Virtual Box.
> 
> It should be a Dell 430 server (or IBM equivalent) CPU Xeon E5-2630 with 4
> 2T SATA HDs/16 G Ram/ RAID PERC H730 controller (supported by mrsas).

That's fine for development or QA/testing environments.

PROD sites should never consist of a single machine, especially one without much
in the way of high-availability features.

> All VMs will run FBSD also with usual Apache/php/Mysql combination.

Apache is fine.  You're (almost?) always bottlenecked by network bandwidth, but for
some purposes nginx + fastcgi is a more suitable choice.

PHP-based e-commerce solutions tend to exhibit the same continuous security flaws
that every other PHP-based system (Wordpress, Joomla, MediaWiki, etc) runs into.

I'd bet that if you talk to your payment processor and see what they recommend, it'd
be Java and not PHP.  Just a guess, but I'm pretty familiar with everyone from
CyberSource, ClearCommerce, AliPay, WorldPay, SagePay, PayPal Payflow/Express, etc
to the smaller regional/country processors like BarclayCard, Garanti, ....

MySQL can be adequate for very low-volume sites, but for read-mostly stuff a no-SQL
database like Mongo or Cassandra are likely to be more performant and for transactional
data something like Postgres or Oracle would be better choices.

> I really would welcome suggestions/opinions on what would be the most
> efficient way to set up those HDs to hold these virtual disks.

You normally shouldn't need to care about disk I/O performance for the app-layer
side, only for the database(s).  Provision your databases on bare metal going by
what the DB recommends for # of spindles to enable separation of tablespace,
temp space, and transaction logs.

Most of the PROD databases I know of working from local storage have heaps of
RAID-1 mirrors, and sometimes larger volumes created as RAID-10 or RAID-50.
Higher volume shops use dedicated SAN filers via redundant Fibre Channel mesh
or similar for their database storage needs.

> I thought about zfs but I won't have lots of RAM avaliable

ZFS wants to be run against bare metal.  I've never seen anyone setup ZFS within
a VM; it consumes far too much memory and it really wants to talk directly to the
hardware for accurate error detection.

> Should I use the controller raid? Gmirror/Graid? What raid level?

Level is easy: a 4-disk machine is suited for either a pair of RAID-1s, a 4-disk RAID-10 volume,
or a 4-disk RAID-5 volume.

Regards,
-- 
-Chuck