port the patch in 1013350 from Bill Darbie (bill-darbie) into latest code base.
incorporate one little comment from Russell about not specifying the default value twice
added a test case
Fixes bug 1013350
Change-Id: Id61f10bf2502fa57d2d26199ba9608e4566e8624
While types other than 'vfat' don't seem to be used at present,
it seems better to allow labels on all file system types,
for future maintenance, and to at least alleviate confusion.
Change-Id: I80110eff577e56e9e48354813efbc3e5488d6257
hypervisor_hostname is set in get_available_resource() and
in get_host_stats(). Except for vmwareapi driver.
Change-Id: I66348ae1e986162b8139b04ab03c42c424338175
When host Compute is down, Nova API deletes the resources associated with the
instance. However, since the instance is marked deleted first, the next query
to update the instance vm_state and task_state in DB fails. And instance
remains in state vm_state=ACTIVE, task_state=DELETING, deleted=True.
Fixes LP: #1074305
Change-Id: I0a4e64d180b4b2c5e398a21a62b29b7d59549a9f
There was a bug where we were checking base image checksums, even
if checksumming was disabled, if there was already a cached checksum
for the image. This resolves bug 1075017.
DocImpact
Change-Id: I38f5914cd9d6326fdf5e5d7f0f57885ff8b5ff32
The poll_rebooting_instances() method for a virt driver can be
called periodically from the manager to handle instances that
may be stuck in a rebooting state. The only information passed
to the virt driver is the configured timeout, which means any
driver implementing this will have to do the same database
query: db.instance_get_all_hung_un_rebooting (..., timeout).
This patch makes the manager do this before calling the virt
driver, passing the appropriate set of instances, which avoids
the need for the driver to make that call.
Related to bp/no-db-compute
Change-Id: I152cb8cd9a107f2d2b1df39401ab7bbe7ff366f1
The default scheduler switched to the filter scheduler, and apparently
the EC2 tests are meant to use the chance scheduler for testing. Also,
compute_scheduler_driver is not recognized when these tests are run
individually.
Fixes bug 1075293
Change-Id: I4a7cae4b3dbd18b8779c44e4c30718a872a82b0a
Update the add_aggregate_host method of the compute rpc api to take in
all aggregate data instead of just the aggregate id. This avoids a db
lookup in the nova-compute service.
Part of blueprint no-db-compute.
Change-Id: Iaccfa28ac23452a4e78581c6b5a6687417283838
Update the finish_revert_resize method of the compute rpc api to take
all of the migration data instead of just the migration id.
This removes the need for doing a db read at the beginning of
finish_revert_resize.
Part of blueprint no-db-compute.
Change-Id: I725404a13408fdd83aef6d19bab6c9481e2eab35
Update the revert_resize method of the compute rpc api to take all of
the migration data instead of just the migration id. This removes the
need for doing a db read at the beginning of revert_resize.
Part of blueprint no-db-compute.
Change-Id: I821cfcbd8edf83a0ffdf64100b6f9224205dea0a
Jenkins