28a1bb3f3c
The charm is documented to take the 'size' in gigabytes. But when passing it down to 'ceph fs subvolume', it's incorrectly calculating the bytes. Closes-Bug: #2078019 Change-Id: I94ebe1bf506ef7741dbf9d2975a7ba82405a41ff Signed-off-by: Ponnuvel Palaniyappan <pponnuvel@gmail.com> |
||
|---|---|---|
| src | ||
| templates | ||
| tests | ||
| unit_tests | ||
| .flake8 | ||
| .gitignore | ||
| .gitreview | ||
| .jujuignore | ||
| .stestr.conf | ||
| .zuul.yaml | ||
| actions.yaml | ||
| build-requirements.txt | ||
| charmcraft.yaml | ||
| config.yaml | ||
| CONTRIBUTING.md | ||
| LICENSE | ||
| metadata.yaml | ||
| osci.yaml | ||
| README.md | ||
| rename.sh | ||
| requirements-dev.txt | ||
| requirements.txt | ||
| test-requirements.txt | ||
| tox.ini | ||
ceph-nfs
Description
CephNFS is a charm designed to enable management of NFS shares backed by CephFS. It supports Ceph Pacific and above.
Usage
CephNFS provides an additional service when deployed with Ceph and CephFS. It should be related to CephMon:
juju add-relation ceph-nfs:ceph-client ceph-mon:client
Once all relations have settled, it is possible to create a new export:
juju run-action --wait ceph-nfs/0 create-share name=test-share size=10 allowed-ips=10.0.0.0/24
The above command has creates an NFS share that is 10GB in size, and is
accessible from any machine in the 10.0.0.0-10.0.0.255 network space. To
grant access to a new network address, the grant-access action should be
used:
juju run-action --wait ceph-nfs/0 grant-access name=test-share allowed-ips=192.168.0.10
This command has granted access to the named share to a specific
address: 192.168.0.1.
It is possible to delete the created share with:
juju run-action --wait ceph-nfs/0 delete-share name=test-share
High Availability
To gain high availability for NFS shares, it is necessary to scale ceph-nfs and relate it to a loadbalancer charm:
juju add-unit ceph-nfs -n 2
juju config vip=10.5.0.100
juju deploy hacluster
juju add-relation ceph-nfs hacluster
Once everything settles, your shares will be accessible over the loadbalancer's vip (10.5.0.100 in this example), and connections will load-balance across backends.
Relations
Ceph-NFS consumes the ceph-client relation from the ceph-mon charm.
Bugs
Please report bugs on Launchpad.
For general charm questions refer to the OpenStack Charm Guide.
Note that starting with the squid track of the ceph-nfs charm, deployment of Ceph Pacific and older clusters is not supported anymore.