229bce2a91
This patch removes the dependency on eventlet.Timeout for cluster data model (CDM) collectors and implements a different timeout handling using futurist waiters, for compute collector. Changes: * Remove eventlet-based timeout wrapper in DecisionEngineSchedulingService - Removed _as_timed_sync_func() and _wrap_collector_sync_with_timeout() - Simplified add_sync_jobs() to call collector.synchronize() directly * Implement timeout handling for concurrent operations in Nova collector: - Add compute_resources_collector_timeout configuration option (default: 600s) - Cancel futures and abort collection gracefully on timeout Assisted-By: Cursor (claude-4.5-sonnet) Change-Id: I7969e1ee295c2c625fbfc1f2acdee13ddd82e7dd Signed-off-by: Douglas Viroel <viroel@gmail.com>
12 lines
580 B
YAML
12 lines
580 B
YAML
---
|
|
upgrade:
|
|
- |
|
|
The previous collector operation timeout mechanism, which was based on the
|
|
`period` configuration option, has been replaced by a new configuration
|
|
option `compute_resources_collector_timeout`. This new configuration
|
|
option provides more explicit control over the timeout for resource
|
|
collection operations in the Compute Cluster Data Model. Operators
|
|
should review and adjust this new configuration according to their
|
|
environment's needs. Note that collectors `period` configuration didn't
|
|
change and continue to work as before.
|