Replies: 1 comment
-
|
Issues on growing memory. Tested 60 running workers with 400 concurrent requests on small flow. [ 1s] RAM 8.93 GB done=0/400 ok=0 err=0 |
Beta Was this translation helpful? Give feedback.
-
|
Issues on growing memory. Tested 60 running workers with 400 concurrent requests on small flow. [ 1s] RAM 8.93 GB done=0/400 ok=0 err=0 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Items to Investigate
The following optimizations should be checked and investigated for potential memory savings:
jemalloc via LD_PRELOAD + MALLOC_CONF
MALLOC_CONFGarbage collection tuning
gc.set_threshold(0, 0, 0)in master process tailgc.freeze()call in worker processesPreload refactoring
setup_llm_cachinginitialization into_run_master_preload_run_master_preloadprestodb/presto#8993
https://instagram-engineering.com/copy-on-write-friendly-python-garbage-collection-ad6ed5233ddf
related merged PRs:
#12587
#12364
#12313
open PRs:
#12778
related:
#12588
Beta Was this translation helpful? Give feedback.
All reactions