feat: Improve OS version filtering for Pub/Sub tasks #5026
Conversation
This commit refines the client-side OS version filtering logic for Pub/Sub tasks to correctly handle scenarios where a legacy bot (without a defined BASE_OS_VERSION) receives a task intended for a specific OS version. Previously, the filtering logic only triggered a skip if both the message and the bot had a defined OS version, and these versions differed. This meant that legacy bots, which do not have BASE_OS_VERSION set, would incorrectly attempt to process tasks explicitly tagged for newer OS versions, leading to errors and inefficiencies. The updated logic now ensures that a task is skipped if: 1. The Pub/Sub message specifies a . 2. The bot's is either different from the message's OS version OR the bot's is not defined (e.g., for legacy bots). This change improves the resilience of ClusterFuzz bots by preventing them from processing incompatible tasks, especially during OS migrations or when mixing legacy and modern bot deployments. Unit tests have been updated to reflect this new behavior, ensuring all scenarios are covered and pass.
|
I wonder if we could have better tested this in our dev/staging environment for #5023 to avoid having these constant hotfixes. |
You're right. In this case, it is genuinely difficult to test since we don't have all behaviors mapped in these environments. The issue actually slipped through an unmapped edge case. |
What kind of legacy bots don't have the variable set and prevented us from catching this case in dev/staging? |
All bots not running on Ubuntu 24.04 are missing this environment variable. As I mentioned, we should have tested more thoroughly, but we missed catching this in our executed tests. |
3 participants
Description:
Summary:
This PR addresses a flaw in the client-side OS version filtering for Pub/Sub tasks. Legacy bots, which do not have the
BASE_OS_VERSIONenvironment variable set, were not correctly skipping tasks intended for specific, newer OS versions. This could lead to bots attempting to execute incompatible tasks, causing errors and wasting resources, particularly during OS version migrations.This change makes the task handling more resilient by ensuring that bots only process tasks compatible with their environment.
Changes:
_filter_task_for_os_mismatch: The core filtering logic insrc/clusterfuzz/_internal/base/tasks/__init__.pyhas been updated. It now correctly skips a task if the incoming message has abase_os_versionattribute and the bot's own OS version is either different or not set (None).src/clusterfuzz/_internal/tests/core/base/tasks/tasks_test.pyhave been updated to validate the corrected logic and ensure there are no regressions.Testing:
All changes are covered by unit tests. The relevant test suite passes successfully.