Skip to content

Enforce project timeout on the analyzer operations

Vishwa Bhat requested to merge listen-sigkill into main

What does this MR do?

There is an issue where some analyzers run beyond the specified project timeout. Shared CI runners run each job under a different process group and when the timeout occurs, the team claims to send a SIGTERM os signal and is followed by a SIGKILL signal after 10 seconds to handle graceful shutdown. Although there are issues(CI or CD process does not receive SIGTERM on termination, another issue comment) stating that the signals are not triggered consistently, ending up running child jobs indefinitely.

However, this MR is an attempt to listen to the SIGTERM signal so that before moving to the subsequent stage(i.e, Analyze, Search, Convert and Serialize) of the run operation, it checks if the SIGTERM signal is received. If yes, then the run job ceases the operation and returns a context-canceled error right away.

What are the relevant issue numbers?

Does this MR meet the acceptance criteria?

Edited by Vishwa Bhat

Merge request reports