Shared/RBAC network support #605
Conversation
# Conflicts: # f5lbaasdriver/v2/bigip/service_builder.py
…sv2-driver into mitaka-m3 # Conflicts: # f5lbaasdriver/v2/bigip/service_builder.py
|
For the is_common_network() change, we need to coordinate Andrew's request with @ssorenso . Steven is working on supporting common networks in the agent. His change there implements is_common_network() based on a setting in the agent INI file, instead of forcing to true as in this patch. Steven's work accomplishes what this patch wants, but gives user control in order to support legacy behavior. |
|
My patch just displays intent, its not indended as a final solution, as we discussed we're more than happy for this to be configurable to support legacy scenarios. There are however a few other fixes included in these changes that make the sync work more consistently and catch a few edge cases, it woud be great if these didn't get lost along the way. |
|
Can one of the admins verify this patch? |
|
Please also disregard the comment from @pjbreaux: This is dust from a set of items we're attempting to do in our background. |
|
Can one of the admins verify this patch? |
1 similar comment
|
Can one of the admins verify this patch? |
…al, because its not handled well in the API but prevents redundant device config.
… neutron_lbaas has already updated to PENDING DLEETE
This reverts commit 7bdfcc8.
|
Can one of the admins verify this patch? |
Added some new RPC functions needed by F5 agent.
|
@abattye - if you'd like to have these merged into master, can you rebase these on top of tree? Also, if you only care about the most recent changes (RPC functions for RBAC network support), can you submit those as a separate pull request? |
* Changes in error handling when objects can't be found for orphan handling and in case scheduling for agent failed * Try 6 times to get an network and don't use bad defaults for the provider segment id and network type in case that the first try fails. Throw exeption if 6 tries fail for the read of network info. * Fix the _get_network_cached method and make it work again. First approach for fixing was completely nonsense * Add network segments to the 30 minute cache for networks. * Enable caching for network segmentation retrieval. Change defaults for network segmentation dict in case not found. * Changes in error handling when objects can't be found for orphan handling and in case scheduling for agent failed * Try 6 times to get an network and don't use bad defaults for the provider segment id and network type in case that the first try fails. Throw exeption if 6 tries fail for the read of network info. * Fix the _get_network_cached method and make it work again. First approach for fixing was completely nonsense * Add network segments to the 30 minute cache for networks. * Enable caching for network segmentation retrieval. Change defaults for network segmentation dict in case not found.
…trieval Fix wrong retry logic for missing segments. Fix bindings assignment in case no agent ist live
…or. Schedule lb's to overloaded agent in case partition is already scheduled. Clean orphan cache every 24 h
…out agent binding)
This contains the necessary driver updates to support the shared network model in this agent pull request F5Networks/f5-openstack-agent#763