You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Multiple Openevolve run calls in parallel result in clobbered loggers #295
If you run openevolve run calls in parallel, each parallel run ends up duplicating each other's logs, because they are all registered against a global logger namespace. This makes the logging basically impossible to use properly
Hi, I'd like to work on this issue. I'll trace how loggers and handlers are set up when two OpenEvolve runs execute concurrently, then add a regression test for duplicated or mixed logs. I'll focus the fix on isolating each run's logging and cleaning up its handlers. The optional log prefix requested in #290 can be a separate follow-up.
Please let me know if there are compatibility constraints you'd like me to account for.
Activity
codelion commentedon Oct 17, 2025
Yeah, this is a known issue, we need to not take the global namespace.
theahura commentedon Oct 18, 2025
I think #297 is related to this. I am seeing system hangs, and when I use pyspy it points to file access contention leading to deadlocks by the logger
Mqw7373 commentedon Sep 29, 2026
Hi, I'd like to work on this issue. I'll trace how loggers and handlers are set up when two OpenEvolve runs execute concurrently, then add a regression test for duplicated or mixed logs. I'll focus the fix on isolating each run's logging and cleaning up its handlers. The optional log prefix requested in #290 can be a separate follow-up.
Please let me know if there are compatibility constraints you'd like me to account for.