Recent changes to ceph check that a directory exists in cephfs before it
will allow an NFS export of the dir to be created. It was a pretty shaky
practice of ours to create NFS exports without dirs backing them anyhow.
This change ensures the dirs exist before the test cases are run and
clean them up afterwards.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Fixes: #815
When updates to the NFS mgr API were recently merged on Ceph main branch
the go-ceph test started failing. This occurred because go-ceph was
unintentionally relying on invalid JSON (an empty string) returned from
the API call for an invalid export pseudo path. On ceph main it now
returns an empty JSON object. This change treats both empty string
and empty JSON object as an error explicitly.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Ceph is adding a flag to get/set NFS-Ganesha's SecType configuration
opiton. This can be used to specify authentication types for an export.
These changes only affect struct fields and are fully backwards
compatible with existing code. Therefore no API status flags are added
to the code.
Test cases added - with api buildtag for ceph main branch.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
The latest ceph pacific should include the mgr changes needed to test
the NFS apis. We should be able to treat nfs like a normal preview api.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
Our test container does not support an orchestrator to pre configure
the ".nfs" configuration pool for us. This is something that rook and
cephadm are expected to do before exports are created, etc.
So we support creating a minimal "mock" configuration in rados.
This option can be toggled on and off via the GO_CEPH_TEST_MOCK_NFS
env var. Turning it off can be useful if you want to run the live
suite against a "real" ceph cluster.
Signed-off-by: John Mulligan <jmulligan@redhat.com>
John Mulligan
Sven Anderson