The --service option to neutron-db-manage is deprecated and will be
removed soon. Use --subproject instead.
Related-Bug: #1501380
Change-Id: I4ab0f7b7c41aeed8aeb14c893f1837db94d23fd3
This changes the instance manager to use the InstanceGroupManager
to manage clusters of instances instead of individual instances,
including clusters of 1 node for non-HA resources.
This also adds some missing documenation on enabling clustering
of the astara-orchestartor service.
Change-Id: Ib679453aafe68e6653c8c5f9f412efa72c2b7cb1
Removes references to Akanda + the rug and replaces them with references
to the actual Astara components. Also drops a bunch of static files that
are not used for anything.
Change-Id: I37c965d39ac07818c88867a011faee35f83026ab
./doc is the standard location for docs across openstack projects.
Move it there to allow us to integrate better with some of the doc
building infra.
Change-Id: I0547301c34e213c154ddb71dc94c10afd9eb8ce2
State machine bits removed as
they are moving to the akanda repo
Aslo fix bad URL in root readme
Partially Implements: blueprint liberty-doc-updates
Change-Id: I35cd688850fdd007bb9f0fb27759f6081e64d160
If calls to boot a router fail (e.g., Nova or Neutron return an HTTP 500 during
the plug + boot process), the state machine goes into CheckBoot state and polls
the router's management address to check for aliveness. This causes a needless
polling process (because the VM will never boot).
Instead, the CreateVM step should make sure that a VM was *actually created*
before proceeding to the CheckBoot step. If not, CreateVM should transition
*back* to CreateVM (up to `reboot_error_threshold` times, until the router is
placed into ERROR state).
When CalcAction determines that the router has an error status, it
transitions to ClearError, which returns to CalcAction. Each time we
make it back to CalcAction, we yield control of the worker and have to
wait for another turn. Since we know we're going to re-compute the same
action, just pass it to ClearError and allow that state to transition
directly to the next real update, bypassing CalcAction.
Change-Id: I5bd28403d679889158f28e6440eab7e3b50318d7
Add a ClearError state to remove the error flag from the VM before doing
anything else with it. Add the new state transition to CalcAction. We
are still protected from boot loops triggered by poll operations because
those are never processed for a router in an error state, but any update
operations or direct commands will be.
Change-Id: I6bf744a33df8f4055a744b5fe2c7ef3d3043c247
Add a command to forcibly rebuild a router. It works
by destroying the instance and building a new one, which is
brute-force, but easy to understand.
I also changed the worker's lookup for tenant router
managers so the tenant id is normalized cleanly to
avoid confusion since some events use '-' and some
do not (yay, consistent!).
Change-Id: I79a622b0c6cc23041da55ec2f54f79ccf463379e
Change the orientation of the state machine diagram so it renders
large enough to read.
Fix reference to CHECKBOOT that should be CHECK_BOOT.
Fix reference to STOP that should be STOP_VM.
Fix order of edge definitions to make them more "logical".