[feature, pod, b1] Allows setting/disabling overcommit of resources (and ratio)

If you look at the capabilities of a virsh pod it has OVER_COMMIT. Which means that MAAS will keep using that Pod, over the available resources.

It currently doesn't support the ability to set the over commit ratio.

Since we can over commit the resources for Virsh Pods, we should set an over commit limit.

I think that the simplest fix here could be enabling/disabling over committing.

thoughts?

+1

On Wed, Jul 26, 2017 at 3:08 PM, Andres Rodriguez <email address hidden>
wrote:

> thoughts?
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1706458
>
> Title:
> MAAS tries to allocate VMs from pods even when resources are exhausted
>
> Status in MAAS:
> Triaged
>
> Bug description:
> MAAS will allocate new VMs from a pod even when resources are
> exhausted:
>
> Here's the output of 'pods read' from my MAAS:
> http://paste.ubuntu.com/25172845/
>
> pod 1 has negative availability numbers,and it keeps on handing out new
> VMs in response to allocate requests:
> "available": {
> "memory": -98772,
> "local_storage": -440440992973,
> "cores": -94
> },
>
> Once resources are exhausted, the allocate request should fail.
>
> This is with 2.2.2 (6094-g78d97d0-0ubuntu1~16.04.1)
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/maas/+bug/1706458/+subscriptions
>

I don't think the correct fix is simply allowing it or disallowing it. I think you should be able to set a ratio, setting a ratio of 1x would be no over committing.

That said, however, it does seem like the default behavior should have
never been to allow unbounded over committing.

Either the default behavior should be to not over commit at all unless the
user specifically requests it.

Otherwise we are simply allowing the usage of a single pod past the
hardware capabilities.

So as a saner short / quick fix we either need to disable overcommitting at
all or have the ability to do it or not. (Again, this is a quick fix.
Setting a over commit ratio is a wishlist).

On Thu, Jul 27, 2017 at 3:21 PM Blake Rouse <email address hidden>
wrote:

> I don't think the correct fix is simply allowing it or disallowing it. I
> think you should be able to set a ratio, setting a ratio of 1x would be
> no over committing.
>
> --
> You received this bug notification because you are subscribed to MAAS.
> https://bugs.launchpad.net/bugs/1706458
>
> Title:
> MAAS tries to allocate VMs from pods even when resources are exhausted
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/maas/+bug/1706458/+subscriptions
>
--
Andres Rodriguez (RoAkSoAx)
Ubuntu Server Developer
MSc. Telecom & Networking
Systems Engineer