Skip to content

fix(monitor): fix wandb internal RSS growth at high sample log frequency #2528

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
mikasenghaas wants to merge 2 commits into
base: fix/orchestrator-rss-growth
Choose a base branch
from
Draft

fix(monitor): fix wandb internal RSS growth at high sample log frequency #2528

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2ad813f
fix(orchestrator): trim glibc heap after each step to prevent RSS growth
mikasenghaas May 17, 2026
4bd0db5
fix(monitor): fix wandb internal RSS growth at high sample log frequency
mikasenghaas May 17, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions src/prime_rl/orchestrator/orchestrator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
import asyncio
import ctypes
import gc
import os
import time
Expand Down Expand Up @@ -797,6 +798,10 @@ def compute_solve_rates(df):
del train_rollouts, train_examples, training_batch, vlm_cache
del results_df, metrics_df
gc.collect()
# Return free glibc heap pages to the OS. numpy/pandas allocate array data
# via malloc (outside Python's allocator), so gc.collect() alone doesn't
# reclaim the RSS. malloc_trim(0) forces glibc to return freed pages.
ctypes.CDLL("libc.so.6").malloc_trim(0)

event_loop_lag_monitor.reset()

Expand Down
Loading