setup environment for external mongo testing

[Ivan Berezovskiy]

I have a blueprint for external mongodb support (https://blueprints.launchpad.net/fuel/+spec/external-mongodb-support). Patches are ready and tested, but we need a testing ci for this feature. It could be a single job that setups mongo in single node or mongo with replicaset depending on environment which we are want to test. If it's a HA mode, we deploy mongo with replicaset, else we deploy only one node.

To deploy mongo we can use our puppet scripts in fuel-library (we need in mongodb module only). To use it we should pass the same parameter to it as it does during fuel deploy. If we deploy HA mode we pass mongodb_bind_address, ceilometer_metering_secret, ceilometer_db_password, ceilometer_replset_members for primary mongo and mongodb_bind_address for another mongo nodes. In simple mode we should deploy only one mongo node as primary.

Then we should pass mongo credentials to job that will test deploy cluster with external mongo.

[Igor Shishkin (teran)]

Ivan, currently we're trying to not to support anything we're not working with directly.
So I can provide you access to our puppet manifests to add/use modules for mongo and help to setup notifications from zabbix for this mongo to you or your team.

But unfortunately we can't support any thing you're developing with.

[Dina Belova (dbelova)]

Igor, the blueprint about external mongodb support has been already approved for lots of reasons - the main is not to pretend maintain and support ALL possible mongo configurations that might be needed by customer. The only moment that prevents these changes to be merged - no test env available. That's why Ivan started working this bug. Actually I don't understand something you've just said - 'I can provide you access to our puppet manifests to add/use modules for mongo and help to setup notifications from zabbix for this mongo to you or your team' - we can't support all mongo configurations that might be needed in real-life deployments in fuel! The easiest way is to add opportunity of external mongodb support here....

[Igor Shishkin (teran)]

Dina,

If we're working on any kind of service that means we should react on it's fails, monitor it and have an SLA for that.
That's pretty bad for aside services which are not involved in infrastructure, which comes from aside and where we don't have enough expertise.

So I suggesting Ivan to write puppet manifests on his own and provide our infrastructure(puppet master, zabbix) as a services to use them for deployment and monitoring.
In this scenario he will be responsible for that service and SLA on his side.
We could only help to order VMs/HW and consult on implementation.

As I've got from Ivan few minutes ago in Skype call you don't have to support all configurations, you need just two of them(Replicaset and single), so it's could be implemented in manifests by Ivan or any of your guys.

[Igor Shishkin (teran)]

https://review.fuel-infra.org/#/c/1259/ will be merged after HCF