Single Storage Pool Cannot be Used by multiple storage endpoints "no available machine matches constraints"

[Dmitrii Shcherbakov]

juju create-storage-pool nvme-osd-devices maas tags=ssd
juju create-storage-pool nvme-bluestore-wal maas tags=optane

In a bundle using this works:

    storage:
      osd-devices: 'nvme-osd-devices'
      bluestore-wal: 'nvme-bluestore-wal'
      #bluestore-db: 'nvme-bluestore-wal'

this does not:
    storage:
      osd-devices: 'nvme-osd-devices'
      bluestore-wal: 'nvme-bluestore-wal'
      bluestore-db: 'nvme-bluestore-wal'

19 down pending xenial No available machine matches constraints: [('interfaces', ['ceph-access:space=6;cloud-compute:space=6;ceph:space=6;mon:space=7;compute-peer:space=6;secrets-storage:space=6;cluster:space=4;ephemeral-backend:space=6;nrpe-external-master:space=6;public:space=7;amqp:space=6;image-service:space=6;0:space=6;lxd:space=6;neutron-plugin:space=6;nova-ceilometer:space=6;shared-db:space=6;internal:space=6']), ('agent_name', ['ab00d651-2f2b-423f-88aa-7d721dbb5ee7']), ('zone', ['default']), ('tags', ['nvme']), ('storage', ['root:0,6:1(optane),7:1(optane),8:1(ssd)'])] (resolved to "interfaces=ceph-access:space=6;cloud-compute:space=6;ceph:space=6;mon:space=7;compute-peer:space=6;secrets-storage:space=6;cluster:space=4;ephemeral-backend:space=6;nrpe-external-master:space=6;public:space=7;amqp:space=6;image-service:space=6;0:space=6;lxd:space=6;neutron-plugin:space=6;nova-ceilometer:space=6;shared-db:space=6;internal:space=6 storage=root:0,6:1(optane),7:1(optane),8:1(ssd) tags=nvme zone=default")

All spaces are there and configured on respective machines (commenting out the last storage binding proved that spaces are irrelevant here):
https://pastebin.canonical.com/p/2t9dbbFG6C/

[Dmitrii Shcherbakov (dmitriis)]

This behavior affects charm-ceph-osd when wal and db need to be collocated.

To work around this issue, db needs to use the wal storage endpoint resolution.

https://review.openstack.org/#/q/topic:fe-collocate-if-db-unspecified+(status:open+OR+status:merged)

[Ian Booth (wallyworld)]

This looks like a maas issue - the error message comes from maas. Juju is simply asking for a machine with 2 storage devices, each tagged "optane"; in this case the default size for each pool would be 1GiB as I don't think the charm specifies a default storage size. The storage constraint passed to maas is:

root:0,6:1(optane),7:1(optane),8:1(ssd)

This corresponds to asking for 1 disk from the osd-devices pool (tagged ssd) and 2 disks from the bluestore pool (tagged optane), and a default sized root disk.

You'd need to look at the maas logs to see why the request could not be satisfied.

[James Page (james-page)]

This is quite an old bug and MAAS + Juju storage is still not a commonly used combination (configuration of devices directly is mostly used rather than storage support).

Is this still and issue?

[James Page (james-page)]

I've marked the charm tasks as invalid as I don't think this is actually a charm issue.

[Launchpad Janitor (janitor)]

[Expired for juju because there has been no activity for 60 days.]