| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Additionally, we drop the "ufo" references from setup.py, spec file
and README, and add the HISTORY file describing how the repo was
initially created. We also update the RPM spec file to use the name
"gluster-for-swift" to avoid colliding with existing RPM names from
RHS 2.0 (the spec file's description was also updated, along with
the version number).
Change-Id: If804224a94208d57896e4189c63736ffc9e01d5e
Signed-off-by: Peter Portante <peter.portante@redhat.com>
Reviewed-on: http://review.gluster.org/4966
Reviewed-by: Luis Pabon <lpabon@redhat.com>
Reviewed-by: Mohammed Junaid <junaid@redhat.com>
Reviewed-by: Kaleb KEITHLEY <kkeithle@redhat.com>
Reviewed-by: Peter Portante <pportant@redhat.com>
Tested-by: Peter Portante <pportant@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Situation:
The function get_part_nodes is being called by Openstack-Swift's proxy/controllers/base.py:
https://github.com/openstack/swift/blob/1.7.4/swift/proxy/controllers/base.py#L410
https://github.com/openstack/swift/blob/1.7.6/swift/proxy/controllers/base.py#L447
As this was not implemented in the current GlusterFS version of ring.py, it was calling swift's
original get_part_nodes, which would often return the incorrect node, resulting in the incorrect
drive being associated with a request.
There is another function that the original ring.py implements -- get_other_nodes, which has to
do with replication. Since GlusterFS is handling replication, this function should never be
called. However, in the interest of completeness, that function is also being replaced.
Code changes:
The two functions, get_part_nodes, and get_other_nodes have been implemented to override the
default functions, and get_nodes has been updated to store information in self vars, about the
account being operated on, as the two new functions are not called with that info, and get_nodes
appears to always be called first.
The code has be refactored to all call _get_part_nodes, much like swift has refactored their code.
Reason for implementation this way:
I didn't see a better way to do it, but am open to suggestions.
Test cases:
A mock ring is created with two different devices, test and iops
test_first_device: Ensure that the first device, test, is returned for both get_nodes, and
get_part_node, and get_more_nodes returns volume_not_in_ring.
test_invalid_device: Ensure that a request for a non-existant device returns volume_not_in_ring.
test_second_device: Same as test_first_device, but for the second device, iops instead of test.
test_second_device_with_reseller_prefix: Test that calling with the reseller prefix, AUTH_ will
still return the correct data.
Change-Id: I2f3d526934a47b01e9c065d0edf0fbf06f300369
BUG: 924792
Signed-off-by: Alex Wheeler <wheelear@gmail.com>
Reviewed-on: http://review.gluster.org/4748
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Kaleb KEITHLEY <kkeithle@redhat.com>
Reviewed-by: Jeff Darcy <jdarcy@redhat.com>
Reviewed-by: Anand Avati <avati@redhat.com>
|
