Installation of octopus/stable charm on Focal fails with default config source=octopus

[Kellen Renshaw]

On a fresh deployment of Focal + Octopus, the ceph-mon charm fails to install the software.

Digging reveals that the charm attempts to use the "focal-ussuri" cloud archive, which doesn't exist.

Manually setting the source to "distro" fixes the issue. The ceph-osd charm does not have this issue.

[Alex Kavanagh (ajkavanagh)]

Please could you provide more details of the install environment and how the charm was installed. Please have a read of https://docs.openstack.org/charm-guide/2023.1/community/software-bug.html and provide some support data.

You mention that you dug into the code. It would be really useful to provide a link and a paste of the code in question.

Thanks.

[Kellen Renshaw (krenshaw)]

The appears to be a result of the default set in config.yaml. Patch attached.

[Kellen Renshaw (krenshaw)]

@Alex, certainly!

Juju 2.9.45 with MAAS 3.4 using a Virsh KVM pod for machines.

Charm was deployed using:
juju model-config series=focal
juju deploy -n 3 ceph-mon --channel octopus/stable --constraints tag=ceph-mon

Then I waited and the "install" hook failed. The requested machines were allocated and built successfully, just the charm install failed.

Workaround that allowed a successful install was "juju config ceph-mon source=distro"

[Kellen Renshaw (krenshaw)]

I should note that the patch is against the stable/octopus branch.

[Felipe Reyes (freyes)]

Hi Kellen,

What's the output of "juju status" and "juju show-application ceph-mon"?

Best,

[Kellen Renshaw (krenshaw)]

juju status:
Model Controller Cloud/Region Version SLA Timestamp
demo lab-default lab/default 2.9.45 unsupported 12:29:07-07:00

App Version Status Scale Charm Channel Rev Exposed Message
ceph-mon 15.2.17 active 3 ceph-mon octopus/stable 177 no Unit is ready and clustered
ceph-osd 15.2.17 active 7 ceph-osd octopus/stable 526 no Unit is ready (1 OSD)
ceph-radosgw 15.2.17 active 1 ceph-radosgw octopus/stable 543 no Unit is ready

Unit Workload Agent Machine Public address Ports Message
ceph-mon/0* active idle 0 192.168.1.40 Unit is ready and clustered
ceph-mon/1 active idle 1 192.168.1.41 Unit is ready and clustered
ceph-mon/2 active idle 2 192.168.1.42 Unit is ready and clustered
ceph-osd/0* active idle 3 192.168.1.43 Unit is ready (1 OSD)
ceph-osd/1 active idle 4 192.168.1.44 Unit is ready (1 OSD)
ceph-osd/2 active idle 5 192.168.1.46 Unit is ready (1 OSD)
ceph-osd/3 active idle 6 192.168.1.48 Unit is ready (1 OSD)
ceph-osd/4 active idle 7 192.168.1.47 Unit is ready (1 OSD)
ceph-osd/5 active idle 8 192.168.1.49 Unit is ready (1 OSD)
ceph-osd/6 active idle 9 192.168.1.45 Unit is ready (1 OSD)
ceph-radosgw/0* active idle 10 192.168.1.50 80/tcp Unit is ready

Machine State Address Inst id Series AZ Message
0 started 192.168.1.40 liked-bug focal default Deployed
1 started 192.168.1.41 large-bird focal default Deployed
2 started 192.168.1.42 great-baboon focal default Deployed
3 started 192.168.1.43 upward-bison focal default Deployed
4 started 192.168.1.44 golden-gnat focal default Deployed
5 started 192.168.1.46 happy-puma focal default Deployed
6 started 192.168.1.48 frank-calf focal default Deployed
7 started 192.168.1.47 fair-swan focal default Deployed
8 started 192.168.1.49 wired-doe focal default Deployed
9 started 192.168.1.45 happy-aphid focal default Deployed
10 started 192.168.1.50 united-cattle focal default Deployed

[Kellen Renshaw (krenshaw)]

$ juju show-application ceph-mon
ceph-mon:
  charm: ceph-mon
  series: focal
  channel: octopus/stable
  constraints:
    arch: amd64
    tags:
    - ceph-mon
  principal: true
  exposed: false
  remote: false
  life: alive
  endpoint-bindings:
    "": alpha
    admin: alpha
    bootstrap-source: alpha
    client: alpha
    cluster: alpha
    dashboard: alpha
    mds: alpha
    mon: alpha
    nrpe-external-master: alpha
    osd: alpha
    prometheus: alpha
    public: alpha
    radosgw: alpha
    rbd-mirror: alpha

[Kellen Renshaw (krenshaw)]

I did work around the issue with setting source=distro, so the status is happy now.

[Felipe Reyes (freyes)]

octopus is ceph-16.x on focal, this predates the openstack-release package introduction, and charm-
helpers doesn't know how to translate "octopus" to "distro" on focal, this works correctly only on
>=wallaby , IIRC we had a bug open for this.

On Fri, 2023-12-01 at 19:31 +0000, Kellen Renshaw wrote:
> I did work around the issue with setting source=distro, so the status is
> happy now.
>

--
Felipe Reyes
Software Engineer @ Canonical
<email address hidden> (GPG:0x9B1FFF39)
Launchpad: ~freyes | IRC: freyes

[Kellen Renshaw (krenshaw)]

Hi Felipe!

Ceph is version 15.X in Focal:
$ rmadison ceph
...
 ceph | 15.2.1-0ubuntu1 | focal | amd64
 ceph | 15.2.17-0ubuntu0.20.04.3 | focal-security | amd64
 ceph | 15.2.17-0ubuntu0.20.04.4 | focal-updates | amd64
 ceph | 15.2.17-0ubuntu0.20.04.5 | focal-proposed | amd64

I don't understand the rest of your comment, but I believe that the config.yaml default is wrong at least.

[Alex Kavanagh (ajkavanagh)]

> I don't understand the rest of your comment, but I believe that the config.yaml default is wrong at least.

Hi Kellen

The default of 'octopus' for the source config value is correct, it's just that the charm isn't handling it correctly on focal. It means that charm-helpers, the library that (almost) all the OpenStack/Ceph/OVN and support charms use from OpenStack engineering, isn't correctly translating 'octopus' to distro on focal along side the ussuri charms. It's probably a bit tricky to sort out, as the openstack-release package didn't exist in the focal release.

The workaround is to (as you've found) pass source=distro when ceph-mon is deployed on focal.