vllm-project / vllm-project/aibrix

storm service failed to detect other roles

Open
#1,345 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Go
Stars
5.1k
Forks
694
Avg merge
1d 19h
Merged PRs (30d)
98

Description

### 🐛 Describe the bug

```
disagg-proxy-router 1/1 Running 0 3m31s
vllm-dev-roleset-xwz6n-decode-5b8b6d795d-0 1/1 Running 0 7m51s
vllm-dev-roleset-xwz6n-prefill-68cf46fbb4-0 1/1 Running 0 7m51s
```

If we put the router into same stormservice a new role, it's hard to detect it using such way. because the ROLE_TEMPLATE_HASH only track the current role instead of other roles. As a result, inter role discovery needs other means
```
--prefiller-host "${ROLESET_NAME}-prefill-${ROLE_TEMPLATE_HASH}-0.${STORM_SERVICE_NAME}.default.svc.cluster.local" \
--prefiller-port 8000 \
--decoder-host "${ROLESET_NAME}-decode-${ROLE_TEMPLATE_HASH}-0.${STORM_SERVICE_NAME}.default.svc.cluster.local" \
--decoder-port 8000
```

### Steps to Reproduce

N/A

### Expected behavior

N/A

### Environment

nightly

Contributor guide

Open the contributing guide

Research direction

Start by tracing the StormService role-discovery path and references to ROLE_TEMPLATE_HASH, then reproduce the multi-role setup shown in the issue. Done means the router can discover the prefill and decode roles when they share a StormService, without relying only on the current role's template hash.

Written by the indexing model from the issue text.

Assessment

Tech stack
go, kubernetes
Domain
backend, distributed-systems
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.