Skip to content
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.

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

Memory leaks on infer requests #2102

Open
darkestpigeon opened this issue Oct 21, 2023 · 1 comment
Open

Memory leaks on infer requests #2102

darkestpigeon opened this issue Oct 21, 2023 · 1 comment
Assignees
Labels
bug Something isn't working

Comments

@darkestpigeon
Copy link

Describe the bug
The server leaks memory on infer requests.

To Reproduce
To reproduce

  1. Get the archive ov-leak-debug.tar.gz, it contains a load simulation script, docker-compose file to run the server, the models and the model_config.json, and a script that creates the models (for reference).
  2. Run the server with docker compose up -d.
  3. Install the dependencies pip install -r requirements.txt.
  4. Run the script python generate_ovms_load.py {model_name} --n-workers 10 --n-threads 10 with {model_name} being static, dynamic or dynamic-nms.
  5. Check OVMS memory usage to see it creep up. It fluctuates for static, creeps up slowly for dynamic, and grows rapidly for dynamic-nms. The memory usage doesn't go down even when the load is no longer applied.

Expected behavior
The memory used by OVMS is constant (or stabilizes after some time).

Logs
Doesn't apply, no explicit errors present.

Configuration

  1. OpenVINO Model Server 2023.1.d789fb785, OpenVINO backend 2023.1.0.12185.9e6b00e51cd
  2. Checked on 12th Gen Intel(R) Core(TM) i9-12900H and on AMD Ryzen 9 5950X 16-Core Processor
  3. Config and models included in the archive
@darkestpigeon darkestpigeon added the bug Something isn't working label Oct 21, 2023
@dkalinowski
Copy link
Collaborator

dkalinowski commented Dec 7, 2023

@darkestpigeon Thank you for your report. We have tested the models and scripts from attachments. This is 4 days (with some breaks) overview (docker container memory usage + RPS):
workload_monitoring

How long was your testing workload?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

2 participants