|
|
@@ -0,0 +1,74 @@
|
|
|
+FAQ
|
|
|
+===
|
|
|
+
|
|
|
+1. Using cloudbridge across zones
|
|
|
+
|
|
|
+ Currently, each instance of a cloudbridge provider is designated to work within a
|
|
|
+ particular zone, for reasons clarified here: :ref:`single-zone-provider`.
|
|
|
+
|
|
|
+ To perform cross-zonal operations, we recommend cloning the provider into a different
|
|
|
+ zone as in this example:
|
|
|
+
|
|
|
+ .. code-block:: python
|
|
|
+
|
|
|
+ all_instances = []
|
|
|
+ for zone in provider.compute.regions.current.zones:
|
|
|
+ new_provider = provider.clone(zone=zone)
|
|
|
+ all_instances.append(list(new_provider.compute.instances))
|
|
|
+ print(all_instances)
|
|
|
+
|
|
|
+
|
|
|
+2. Cleaning up resources/left over resources
|
|
|
+
|
|
|
+ The trickiest part about using cloud resources is the orderly cleanup of resources
|
|
|
+ when they are no longer needed. Cleanup is often complicated, as cloud-providers
|
|
|
+ may have delays in responding at certain times, and transient errors at other times.
|
|
|
+ While cloudbridge does not designate a particular strategy to combat this,
|
|
|
+ the `controller pattern`_ is a recommended mechanism for handling such scenarios:
|
|
|
+
|
|
|
+
|
|
|
+ Cloudbridge provides some utilities that can aid in simpler scenarios, such as
|
|
|
+ `wait_for`, the cleanup helper and retries.
|
|
|
+
|
|
|
+ The following example demonstrates a scenario where an instance and its attached
|
|
|
+ volume must be deleted.
|
|
|
+
|
|
|
+ .. code-block:: python
|
|
|
+
|
|
|
+ from cloudbridge.base import helpers as cb_helpers
|
|
|
+ import tenacity
|
|
|
+
|
|
|
+ def does_instance_or_volume_still_exist(inst, vol):
|
|
|
+ return provider.compute.instances.get(inst.id) or
|
|
|
+ provider.storage.volumes.get(vol.id)
|
|
|
+
|
|
|
+ def detach_and_delete(inst, vol)
|
|
|
+ with cb_helpers.cleanup_action(lambda: inst.delete()):
|
|
|
+ vol.detach()
|
|
|
+ vol.wait_for(
|
|
|
+ [VolumeState.AVAILABLE],
|
|
|
+ terminal_states=[VolumeState.ERROR, VolumeState.DELETED])
|
|
|
+ vol.delete()
|
|
|
+ self.wait_for([VolumeState.UNKNOWN, VolumeState.ERROR])
|
|
|
+
|
|
|
+ def delete_my_instance_and_attached_volume(provider, instance, vol):
|
|
|
+ retryer = tenacity.Retrying(
|
|
|
+ stop=tenacity.stop_after_delay(300),
|
|
|
+ retry=tenacity.retry_if_result(does_instance_or_volume_still_exist(instance, vol),
|
|
|
+ wait=tenacity.wait_fixed(5))
|
|
|
+
|
|
|
+ retryer(detach_and_delete, instance, vol)
|
|
|
+
|
|
|
+ # invoke with the instance and vol you want to delete
|
|
|
+ delete_my_instance_and_attached_volume(my_inst, my_vol)
|
|
|
+
|
|
|
+
|
|
|
+ The code above attempts to first detach and then delete the volume.
|
|
|
+ If an exception occurs, such as the volume not existing, the `cleanup_action` code
|
|
|
+ ensures that the `inst.delete()` code runs regardless of the success or failure
|
|
|
+ of the volume deletion operation. The tenacity.retryer wraps the entire operation
|
|
|
+ so that the overall process will repeat till both the volume nor the instance no
|
|
|
+ longer exist.
|
|
|
+
|
|
|
+
|
|
|
+.. _controller pattern: https://kubernetes.io/docs/concepts/architecture/controller/#controller-pattern
|
Nuwan Goonasekera