There are duplicated host bucket in ceph osd tree

[Jeffrey Zhang]

Deploy kolla rocky with ceph enabled. after it is done, you will found there are two same node bucket in ceph osd tree like below

ID CLASS WEIGHT TYPE NAME STATUS REWEIGHT PRI-AFF
-1 1.00000 root default
-2 1.00000 host 10.3.0.21
-2 1.00000 host node1
 0 hdd 1.00000 osd.0 up 1.00000 1.00000

the 10.3.0.21 and node1 are the same node.

The reason come from this[0] patch. The `osd crush update on start` logical is moved from `ceph-osd-prestart.sh` to ceph-osd startup process. So ceph-osd will create `host node1` bucket automatically. Whereas, kolla is creating `host 10.3.0.21` from the `osd/extend_start.sh` script.

For the less confused and ceph upgrade impact, i think we should disable `osd crush update on start` feature in default.

[0] https://github.com/ceph/ceph/commit/a28b71e3c9446541e14795324f2ec1f9d69c9187

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/615483
Committed: https://git.openstack.org/cgit/openstack/kolla-ansible/commit/?id=6db3f9f3422400e408136501b817ded3cddc5505
Submitter: Zuul
Branch: master

commit 6db3f9f3422400e408136501b817ded3cddc5505
Author: Jeffrey Zhang <email address hidden>
Date: Mon Nov 5 15:02:40 2018 +0800

    Disable ceph osd crush update on start in default

    The buggy come from ceph changes[0], which is included since ceph osd
    v11.0.0. The `osd crush update on start` logical is moved from
    `ceph-osd-prestart.sh` to ceph-osd startup process. So ceph-osd will
    create buckets by node hostname automatically. Whereas, kolla is
    creating buckets by node ip

    For the less confused and ceph upgrade impact, disable `osd crush update
    on start` is a better choice

    [0] https://github.com/ceph/ceph/commit/a28b71e3c9446541e14795324f2ec1f9d69c9187

    Change-Id: Ibbeac9505c9957319126267dbe6bd7a2cac11f0c
    Closes-Bug: #1801662

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/kolla-ansible 8.0.0.0b1 development milestone.