manila/releasenotes/notes/bp-update-cephfs-drivers-9ac5165f31669030.yaml
Victoria Martinez de la Cruz a830710939 Update cephfs drivers to use ceph-mgr client
Use python rados client to talk to the ceph-mgr service.

A python rados client is created by the driver that lasts
during the driver's lifecycle.

The drivers can now work with multiple filesystem clusters.
The filesystem to be used by manila can be specified by the
driver option 'cephfs_filesystem_name'.

The removal of a share will be quicker for the manila user.
The ceph-mgr volumes module moves the share's content to
a trash folder and purges the trash's contents
(`rm -rf` of the backend CephFS subvolume/subtree) aysnchronously,
whereas the ceph_volume_client library moves the share's content
and purges the content synchronously.

Implements: bp update-cephfs-drivers

Co-Authored-By: Victoria Martinez de la Cruz <victoria@redhat.com>
Co-Authored-By: Ramana Raja <rraja@redhat.com>
Co-Authored-By: Tom Barron <tpb@dyncloud.net>

DocImpact

Change-Id: I1f81db1ba7724c0784d87f9cb92bb696f6778806
2021-03-24 10:50:05 -04:00

37 lines
1.8 KiB
YAML

---
deprecations:
- |
As of the Wallaby release the CephFS driver no longer recognizes
the scoped extra-spec ``cephfs:data_isolated`` because it is no
longer supported by the Ceph community. This style of data isolation
required dedicating a Ceph pool for each share and scaled and performed
poorly.
- |
The ``ceph_volume_client`` is deprecated by the CephFS driver in favor of
a python rados client that connects to the Ceph manager daemon to interact
with the Ceph cluster. This new connection method will enable functionality
not available with older client, which has been deprecated by the Ceph
community and will be removed in the Quincy release.
upgrade:
- |
Manila's CephFS drivers now **require** the "python3-ceph-argparse" and
"python3-rados" packages. Do not upgrade without adding these packages
to the environment where the ``manila-share`` service runs since
without them the driver will refuse to start up. This breaking change
is necessary because the old ``ceph_volume_client`` has been deprecated
by the Ceph community.
features:
- |
The Ceph backend can now work with multiple filesystem clusters.
The filesystem to be used by manila can be specified by the
driver option 'cephfs_filesystem_name'. If this option is not specified,
the driver will assume that a single filesystem is present in the Ceph
cluster and will attempt to use it.
- |
Deletion of shares offerd by the CephFS driver (CephFS and NFS) is
now faster. Now the Ceph manager moves deleted share's content to a
trash folder and purges the contents asynchronously rather than
handling this as part of the synchronous delete operation. The purge
can take considerable time if a share contains a significant amount of data.