There's a new driver mode supported since
Antelope (introduced in Zed) that supports
ceph orchestrator deployed CephFS NFS clusters.
We were missing documentation regarding this.
Also consolidate the admin/configuration references
for this driver; this deletes all the outdated
information in the existing config guide.
Partially-implements: bp use-cephadm-nfs-ganesha
Change-Id: I43952d890629e60465059a5b4242817becb198b1
Signed-off-by: Goutham Pacha Ravi <gouthampravi@gmail.com>
Deprecate the create snapshot from share group feature
in CephFS drivers (both Native and NFS Ganesha).
This needs to be done since the subvolume group snapshot
feature is no longer supported in mainline CephFS, meaning that,
if we leave the call in place, an exception will be raised
from the backend and lead to an error status for the attempted
snapshot from share group.
Existing group snapshots can still be listed and deleted, so these
operations stay the same.
Closes-Bug: #1925332
Change-Id: I49efb7042d38257416875f40c03923e974b2eb8e
As part of our work on updating CephFS drivers
to use the ceph-mgr interface, the related docs
on the admin guide need to be updated
- Updates ceph caps for manila to include mgr caps,
remove mds caps, and constrain mon caps
- Add the new cephfs_filesystem_name param to the
config
- Add upgrade to Wallaby considerations (minimum Ceph
version required)
Partially-Implement: bp update-cephfs-drivers
Change-Id: I113639cc9989bbaf224d19969993870d30b81c29
Adds a missing cap to access mgr and be able to list subvolumegroups
and subvolumes in a cephfs volume
Change-Id: Ie9849ed875b996cf33aeb296f258ec84c8b720ef
Closes-Bug: #1907467
Include Wallaby (master) support matrix and
some extra considerations to consider on manila
with cephfs backends.
Change-Id: Iad7cb229151b8dd707a59553cb6e966090d0eeae
The driver inflated total and available capacity
due to an incorrect calculation. The driver was
also ignoring the configuration option
"reserved_share_percentage" that allows
deployers to set aside space from scheduling
to prevent oversubscription.
While this bugfix may have an upgrade impact,
some things must be clarified:
- Inflating the total, free space will allow
manila to schedule workloads that may run
out of space - this may cause end user
downtime and frustration, because shares are
created (empty subvolumes on ceph occupy no
space) easily, but they could get throttled
as they start to fill up.
- CephFS shares are always thinly provisioned
but, the driver does not support oversubscription
via manila. So, real free space is what
determines capacity based scheduler decisions.
Users however expect share sizes to be honored,
and manila will allow provisioning as long
as there is free space on the cluster. This
means that Ceph cluster administrators
must manage oversubscription outside of manila
to prevent misbehavior.
Depends-On: Ic96b65d2caab788afca8bfc45575f3c05dc88008
Change-Id: I6ab157d6d099fe910ec1d90193783b55053ce8f6
Closes-Bug: #1890833
Signed-off-by: Goutham Pacha Ravi <gouthampravi@gmail.com>
The CephFS driver optionally advertised snapshot
support to manila, based on a configuration
option "cephfs_enable_snapshots". Manila does not
provide snapshots by default to users, unless an
administrator chooses to set the "snapshot_support"
extra-spec in the share type they're using. So this
configuration option is really an unnecessary safeguard
at this point.
Moreover, CephFS snapshots are fully supported from
Ceph Nautilus, and with Id847b1b5e6623172f2600afd69e51182067444cb,
manila added a support statement for the Victoria
release that indicates that Nautilus is the minimum
version of ceph supported. So let's default this
option to True, and deprecate this option at the same
time to eliminate this duplicate safeguard.
Partially-implements: bp update-cephfs-drivers
Depends-On: I885ff78586540d01590f98cca2a00d55161dbe04
Change-Id: I1ef42b414f8af0b8f9a912655757ec200a75c309
Signed-off-by: Goutham Pacha Ravi <gouthampravi@gmail.com>
The manila discussed cephfs futures
during the Victoria cycle PTG [1]. To support newer
features in ceph, we'll need to start using the
ceph-mgr interface that debuted in Nautilus.
When we start adopting ceph-mgr, we can no longer
support Mimic, Luminous and older releases of Ceph
without maintaining code workarounds and a continuous
integration system to ensure they're not broken.
This takes considerable effort from community members.
We've supported the Luminous release since OpenStack
Queens, and there have been at least
three other major releases of Ceph in the meantime,
let's add a release note and adjust the ceph driver
documentation calling out our support/testing stance.
[1] http://lists.openstack.org/pipermail/openstack-discuss/2020-June/015494.html
Partially-implements: bp update-cephfs-drivers
Change-Id: Id847b1b5e6623172f2600afd69e51182067444cb
Signed-off-by: Goutham Pacha Ravi <gouthampravi@gmail.com>
Allow the ganesha driver to store ganesha exports
and export counter as Ceph RADOS objects. This enables
highly available(HA) Ganesha servers in manila deployments
to store their config in a HA storage.
Implements: blueprint ganesha-ha-rados
Change-Id: Ia51156055fa10d0661e662c9c998829864f1a204
As agreed in weekly manila community meetings and at the
Queens PTG, several documents currently presented to contributors
are really of general interest to cloud administrators and should
be presented accordingly.
Move the docs on manila back ends, the feature support mapping,
capabilities and extra specs, and export locations and metatdata
to the admin section.
Change-Id: I6d9dec56890fb3a9bc0539c45c43233517cea048