distcloud/distributedcloud/ocf
Kyle MacLeod fc5afdacf2 Refactor portion of dcmanager into dcmanager-state process
This change breaks down the monolithic 'dcmanager' process into a
separate state-handling multi-process architecture. Subcloud state
management is handled by the new state processes. This is a performance
optimization, increasing the throughput of subcloud state change
operations during dcmanager audits. It also optimizes performance
handling of state updates from dcorch.

Refactoring. These top-level methods are moved from dcmanager
SubcloudManager to SubcloudStateManager (along with associated
RPC/service APIs above):
- update_subcloud_endpoint_status
- update_subcloud_availability
- plus internal methods, including the subcloud-level synchronization
  block

New service: DCManagerStateService()
- distributedcloud/dcmanager/state/service.py
    - overriding max_pool_size, max_overflow values from
      /etc/dcmanager/dcmanager.conf

New state manager: SubcloudStateManager()
- distributedcloud/dcmanager/state/subcloud_state_manager.py

New RPC client: SubcloudStateClient()
- Added to existing distributedcloud/dcmanager/rpc/client.py

New RPC topic:
- TOPIC_DC_MANAGER_STATE = "dcmanager-state"

SM service changes:
- distributedcloud/dcmanager/cmd/state.py
- distributedcloud/ocf/dcmanager-state

Other notable changes (from testing):
- Cleanup: unused ThreadGroupManager removed from dcmananager service.py
- generic_sync_manager: Add eventlet yield during subcloud processing
  during initialization
- dcorch: fix exceptions on shutdown due to race on threadgroup shutdown
- dcorch: log service startup with worker config

Test cases:
- update test cases where necessary to used the newly refactored code
- since there is no new logic, no additional tests are required

Test Plan:
PASS:
- Test various audit scenarios in small virtualized environment
- Test audit scenarios in lab setting with a large number of subclouds
    - subclouds going offline/online (including flooding)
    - dead office recovery
    - swact
    - system soak
- Validate dcmanager-state service lifecycle and dependencies

Story: 2009725
Task: 44317

Change-Id: I2c9a0f84e8cf638632ca319545e9e93e6f43f263
Signed-off-by: Kyle MacLeod <kyle.macleod@windriver.com>
2022-01-27 16:33:41 -05:00
..
dcdbsync-api Move content to subdir to support relocated packaging 2019-11-04 13:57:02 -05:00
dcmanager-api Move content to subdir to support relocated packaging 2019-11-04 13:57:02 -05:00
dcmanager-audit Move subcloud audit to separate process 2020-05-14 09:34:23 -05:00
dcmanager-audit-worker Move subcloud audit to new worker processes 2021-02-25 19:04:01 +00:00
dcmanager-manager Set up /var/run/dcmanager as dcmanager's default temp files location 2020-06-17 10:40:23 -04:00
dcmanager-orchestrator Move dcmanager orchestration to a separate process 2020-09-14 11:17:06 -04:00
dcmanager-state Refactor portion of dcmanager into dcmanager-state process 2022-01-27 16:33:41 -05:00
dcorch-cinder-api-proxy Move content to subdir to support relocated packaging 2019-11-04 13:57:02 -05:00
dcorch-engine Move content to subdir to support relocated packaging 2019-11-04 13:57:02 -05:00
dcorch-identity-api-proxy Move content to subdir to support relocated packaging 2019-11-04 13:57:02 -05:00
dcorch-neutron-api-proxy Move content to subdir to support relocated packaging 2019-11-04 13:57:02 -05:00
dcorch-nova-api-proxy Move content to subdir to support relocated packaging 2019-11-04 13:57:02 -05:00
dcorch-patch-api-proxy Move content to subdir to support relocated packaging 2019-11-04 13:57:02 -05:00
dcorch-sysinv-api-proxy Extend sysinv api proxy to support load operations 2020-06-01 16:16:20 -04:00