[ale] FW: SUSE Linux Days 2013 kicks off this May!
Chris Ricker
chris.ricker at gmail.com
Fri May 17 09:02:12 EDT 2013
On 5/15/13 6:57 AM, JD wrote:
> About a year ago, we had a speaker from a company who had deployed over 100
> physical nodes in a lab with OpenStack to see how it worked and learn about it.
> Since that time, they've gone into production. The Atlanta Openstack community
> has monthly meetings ... there is a meetup group. I think they meet at Manuel's
> Tavern. BlueHost was well represented.
Meetups are at Manuels @ 7 pm on the 3rd Thursday of every month so they
conflict with ALE
> What I took from that talk and a few other conversations with the speaker - he
> is a friend - is that for less than 50 nodes, the overhead for any "cloud
> deployment tool" is just too much. It is best to stay with simple and straight
> forward SAN+Net+hypervisor deployments then use something like Ansible or Salt
> to manage setup and breakdowns.
I'd say it's not so much a matter of scale as it is velocity and
operating styles. Deploying a private cloud gives you more speed /
flexibility of deployment than traditional SAN+hypervisor but there is
an overhead associated with changing to cloud that may not be
justifiable if you don't need that, or can't adjust your IT governance
to align with it
> Another friend at the same company doing Openstack has told me about their
> puppet deployment. Seems they followed what a vendor suggested ... sorta like
> "nobody gets fired for buying IBM" and are not pleased. They had problems
> without puppet, now they have problems due to puppet that are only slightly
> less. They are looking at Ansible and SaltStack, but at the scale they work, any
> tool will be difficult to be happy using. It will be hard to switch from the
> market leader due to politics.
Most people running OpenStack in production are using puppet or chef,
with the community roughly evenly split between the two. Some are using
Salt either instead of or in conjunction with one of the above (puppet
for config management + salt for "I need to log into all my nodes and do
ad-hoc command X" is pretty common). Ansible doesn't have much traction
yet. I'd guess after AnsibleWorks gets a bit more established they may
look into supporting OpenStack more aggressively...
If you look at the various vendors' "distros" of OpenStack, some are
open source and very transparently using puppet / chef in an extensible
customizable manner. Others go down the route of trying to make
OpenStack more of an appliance (but even those are typically using
puppet or chef under the covers, they just don't let you touch it)
More information about the Ale
mailing list