puppet-ceph does not configure devices like /dev/nvme0n1 as OSDs

[John Fulton]

NVMe SSDs appear as /dev/nvme0n1 and their partitions appear as /dev/nvme0n1p1, /dev/nvme0n1p2 and so on.
puppet-ceph is fine with such devices being used as Ceph Journals but when they are passed as arguments to be OSDs, then ceph-disk prepare runs more than once and thte deploy fails.

This seems to be because puppet-ceph assumes partition numbers follow the device name but this device this is a device name which ends with a number and then uses pX as its suffix.

puppet-ceph should be updated to handle these types of devices as OSDs

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/462991

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/462991
Committed: https://git.openstack.org/cgit/openstack/puppet-ceph/commit/?id=0754aecaa9cddfe5a6bf7c9bed43c18b35db6508
Submitter: Jenkins
Branch: master

commit 0754aecaa9cddfe5a6bf7c9bed43c18b35db6508
Author: John Fulton <email address hidden>
Date: Mon May 1 13:37:24 2017 -0400

    Correctly configure /dev/nvme0n1 devices as OSDs

    Change pupppet-ceph's assumption that all devices which end with
    a "1" have a first partition so that it can correctly configure
    NVMe SSDs, which appear as /dev/nvme0n1 and whose partitions appear
    as /dev/nvme0n1p1, /dev/nvme0n1p2 and so on.

    Closes-Bug: #1687114
    Change-Id: I433fd3ab3ce240c661e7e9faeb758fd0d0215f6a

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/puppet-ceph 2.4.0 release.