Looking at PaaS.
Passed on CloudFoundry as I do not want to install Ruby which seems to be needed to use that.
So I am looking at OpenShift Origin instead. That works better for me as it’s Redhat oriented and I’m using Fedora. Also I already have OpenStack cloud infrastructure running and I may be able to use OpenShift against that. I am assuming it needs OpenStack… the doc doesn’t list that or Docker as a pre-requisite but I’m guessing it needs somewhere to start a machine instance running.
Will probably get seriously into that over the weekend.
Starting off with downloading the VM images for a test VM (yes images !), I only want a KVM image but both the KVM (VMware) and VirtualBox images are bundled in a single zip file so no choice but to download the entire zip file then remove the VirtualBox image… download will take a while. Install documentation says to convert the VMware image to ‘raw’ format to get a image usable under KVM.
The install documentation for the VM covers installing it using the virt-manager GUI. Not much use on a headless server, but no problems as I am comfortable in building (or rebuilding) VMs from the command line using existing disk images anyway.
One curious thing is the documentation says KVM needs a ‘raw’ disk. I’m assuming that is an OpenShift requirement as it is certainly not a KVM requirement, I have lots of KVM systems running off qcow2 disks without any problems.
Puppet rules to install OpenShift Origin are documented on the OpenShift site. So while I have never played with Puppet it seems it should be trivial to install everythong on a physical machine… but I will not install it on a physical machine as
- It requires MongoDB; which seems a strange choice as Fedora uses postgress as the preferred DB (at least packaged utilities like bacula default to postgress now where they used to use mysql, which is a pain). Anyway I’m mariadb on physical hardware; not interested in installing multiple databases
- And also needs Ruby, which I will not install on a physical machine even if hell does freeze over. Nothing wrong with Ruby, lots of sites use it without problem. I personally have never been able to do a sucessfull Fedora OS upgrade to a new release on any machine with Ruby installed, Fedora packaging conflicts I suppose, but I won’t use it on a physical machine anymore
- And is using ActiveMQ instead of RabbitMQ… which is not a problem but I haven’t come across ActiveMQ before, so will need to go off on a tangent to see what that involves
As an aside annoyingly some of the links on the openshift page (at least the backlog link for the user interface) jump to ‘trello’ which disables the browser ‘back’ button to leave the trello site… which is a pain when trying to check out the documentation links.
However
My openstack test system has custom built images to create web server and database server instances already. I guess what my play with this will determine is if it is more that just another pretty frontend… which (assuming again) I guess it must be as the documentation for the downloaded VM images says 1Gb of memory for the downloaded VM image is fine… so where does the OpenShift VM spin up it’s instances ?… will try spinning the first one up with the router unplugged to make sure it’s not dependant upon external sites. if it needs openstack I want it to use my machines not external ones.
But thats in the weekend… VM disk images still downloading, will be a while.
