Completely fails to honour constraints

[Aaron Bentley]

Using juju 1.16.6 and the ec2 provider:

If I specify a root disk of 10 G and mem of 8G, I get a root disk of 8G (less than 10) and mem of 1.7 G (less than 8).

In order to get an m1.large, I tried specifying 7G mem and 420G root-disk, but this resulted in the same instance type being deployed.

$ juju deploy jenkins --constraints "root-disk=10G mem=8G"
$ juju status
environment: juju-ci3
machines:
  "0":
    agent-state: started
    agent-version: 1.16.6
    dns-name: ec2-54-84-244-38.compute-1.amazonaws.com
    instance-id: i-59745278
    instance-state: running
    series: precise
    hardware: arch=amd64 cpu-cores=1 cpu-power=200 mem=3840M root-disk=8192M
  "1":
    agent-state: started
    agent-version: 1.16.6
    dns-name: ec2-54-84-242-24.compute-1.amazonaws.com
    instance-id: i-06fce526
    instance-state: running
    series: precise
    hardware: arch=amd64 cpu-cores=2 cpu-power=500 mem=1740M root-disk=8192M
services:
  jenkins:
    charm: cs:precise/jenkins-9
    exposed: false
    units:
      jenkins/0:
        agent-state: started
        agent-version: 1.16.6
        machine: "1"
        open-ports:
        - 8080/tcp
        public-address: 172.31.44.6

[Martin Packman (gz)]

The root-disk was not supported in 1.16 and there is a known bug that an invalid constraint causes all constraints to be ignored and the defaults used instead.

Try not specifuing root-disk and the mem constraint should then be obeyed.

[Aaron Bentley (abentley)]

root-disk has been documented for months, and there is no indication that it is not supported: https://juju.ubuntu.com/docs/reference- constraints.html

Even William Reade recommends using it:
http://askubuntu.com/questions/425710/how-can-i-determine-the-available-amount-of-disk-space-when-using-a-charm/427029#427029

We have been using it since at least December, but I believe we started using it much earlier.

On Canonistack, with no root-disk constraint, we get 2.2 G, and with the constraint we get 9.9G.

Do you have evidence that this is *not* a supported feature? AFAICT, it *is* supported.

[John A Meinel (jameinel)]

It has been in trunk for a long time. But you were using 1.16.6 which does
not have it.

John
=:->
On Mar 3, 2014 10:50 PM, "Aaron Bentley" <email address hidden>
wrote:

> root-disk has been documented for months, and there is no indication
> that it is not supported: https://juju.ubuntu.com/docs/reference-
> constraints.html
>
> Even William Reade recommends using it:
>
> http://askubuntu.com/questions/425710/how-can-i-determine-the-available-amount-of-disk-space-when-using-a-charm/427029#427029
>
> We have been using it since at least December, but I believe we started
> using it much earlier.
>
> On Canonistack, with no root-disk constraint, we get 2.2 G, and with the
> constraint we get 9.9G.
>
> Do you have evidence that this is *not* a supported feature? AFAICT, it
> *is* supported.
>
> --
> You received this bug notification because you are subscribed to juju-
> core.
> https://bugs.launchpad.net/bugs/1285343
>
> Title:
> Completely fails to honour constraints
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/juju-core/+bug/1285343/+subscriptions
>

[Aaron Bentley (abentley)]

1.16.6 does have it. As I showed, I get 2.2 G without it, and 9.9 G with it on Canonistack. If the constraint wasn't in stable, I would get an error like this:
~$ juju bootstrap --constraints "rot-disk=10G mem=8G"
error: invalid value "rot-disk=10G mem=8G" for flag --constraints: unknown constraint "rot-disk"

"root-disk" is accepted by 1.16.6 and earlier, and on Canonistack it has the intended effect.

[Martin Packman (gz)]

Did you verify on the machine if it really had the lower root-disk and memory values juju reported? This may be bug 1183259 with nova reporting flavors incorrectly when deletes are involved.

[Aaron Bentley (abentley)]

I got the lower values with the ec2 provider, so I don't think a nova bug could be the cause.

[Martin Packman (gz)]

Geh, should have seen that. Okay, so r1972 added root-disk support for EC2, which is not in the 1.16 branch.