[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Just containers and cartridges

----- Original Message -----
> From: "David Strauss" <david davidstrauss net>
> To: users lists openshift redhat com
> Sent: Friday, June 7, 2013 5:42:46 PM
> Subject: Just containers and cartridges
> Is there a straightforward way to do one of the following?
>  * Set up just the node parts of OpenShift and administer containers
> using a CLI or machine-local API.

There are command-line tools on the node host for doing just about anything the broker can command. I don't think there's any particular documentation of this, and the man pages are pretty skimpy. 

Something I wrote might help you get started: http://sosiouxme.wordpress.com/2013/04/13/openshift-v2-cartridges-node-host-tools/

With no broker in the picture, certain things won't work: app scaling and Jenkins builds. That may not matter to you.

>  * Set up the broker and node parts on the same machine without the
> DNSSEC complexity.

Not sure what you mean by "DNSSEC complexity". Dynamic DNS complexity, maybe.

Dynamic DNS is used because the app's hostname is one of the fundamental routing mechanisms. When an HTTP request comes into the node, the Host: header is used to route requests to the correct gear. So without that, how are you going to distinguish external connections to the various apps running on a node?

There is an answer, actually: cartridges can and mostly do request unique ports on the node host. The current (hard-coded) algorithm allocates 5 ports per gear based on the UID of the gear. Look into where that happens and you can probably find a way to hook into the process for your own management system. 

Another answer is ssh port-forwarding via the gear user. I would consider this less desirable since ssh tunnels kind of require their own management, as they are somewhat fragile and can go down more or less at random. However it is much more flexible.

You asked about setting up a broker and node on the same host. You certainly can (that's the default demo mode). If you just don't want to deal with DNS, maybe the easiest way is to just swap in a do-nothing (or do-what-you-like) DNS plugin. That should be pretty simple. The Dynamic DNS entries are really only used for external routing to the gears. If you don't need that, not creating DNS entries isn't going to hurt you.

> I'd like to explore moving more of my company's container system to
> OpenShift's implementation, but we'd like to isolate the deployment to
> individual container hosts.

You'll be replicating a lot of work that's gone into the broker-based management system. Not saying that won't be valuable, just that it's likely to be a lot of work. But let us know how it goes if you choose that route! Never know where open source innovation will lead, that's the cool thing about being open.

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]