lablup / lablup/backend.ai

Limit the number of sessions (and their kernels) created at a single scheduler tick

Open
#2,952 0 comments 0 reactions 1 assignee Claimed by @fregataa View on GitHub
Dominant language
Python
Stars
670
Forks
183
Avg merge
17h 7m
Merged PRs (30d)
358

Description

When there are a large number of scheduled sessions for a single scheduler tick, making all RPC calls at once may cause the manager performance issues and potential hangs.

Let's apply the concept of [more_itertools.chunked()](https://more-itertools.readthedocs.io/en/stable/api.html#more_itertools.chunked) when making the RPC calls for each kernel creation, introducing a new configuration option to specify the chunk size.

We should also check if the hang happens when there are overlapping RPC calls to the _same_ agent, even when there is a low number of kernels to create.

JIRA Issue: BA-78

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.