Skip to main content

6 posts tagged with "Bug修复"

View all tags

Docker Compose service won't come back after restart? Check the restart policy

· 5 min read
CCLEE
CCLEE

Debugging a Milvus-dependent service that failed to start in a RAG knowledge base project — full writeup below.

TL;DR​

After a host reboot (or a container crash), a group of services didn't come back: the app port had no listener and docker ps -a showed everything Exited. The root cause: docker-compose.yml had no restart policy (default no), so once a container died it stayed dead. Fix: set restart: always on every production service so the infrastructure self-heals after a crash or reboot.

Milvus collection name returns 500? UUID hyphens break the naming rules

· 4 min read
CCLEE
CCLEE

Debugging multi-tenant collection naming in a RAG knowledge base project — full writeup below.

TL;DR​

Concatenating f"{tenant_id}_{collection}" to name a Milvus collection breaks when tenant_id is a UUID: the result starts with a digit and contains hyphens -, both of which violate Milvus naming rules and throw code=1100. The rule in one line: first character must be a letter or underscore, only [a-zA-Z0-9_] allowed, no hyphens. Never concatenate a UUID raw — use the original collection name, or convert it to a legal identifier.

Python task marked failed but no error? try/except swallowed the exception

· 5 min read
CCLEE
CCLEE

Debugging a silent failure where a document sync task marked everything failed in a RAG knowledge base project — full writeup below.

TL;DR​

A shared method was refactored with a new parameter signature, but one caller was missed. The caller passed arguments under the old contract and threw TypeError — except the call sat inside a try/except that quietly funneled the exception into a failed counter. No crash, no ERROR in the logs, just a number ticking up. These "silent failures" are the hardest bugs to track down. Two fixes: grep all callers after a signature refactor; and make except blocks log or re-raise, never swallow silently.

修复 Milvus 混合检索 RRF 分数与相似度阈值不兼容

· 3 min read
CCLEE
CCLEE

在 RAG 知识库项目中调试混合检索评分问题，以下是完整排查过程。

TL;DR​

Milvus 混合检索的加权融合分数 = 0.7 * dense_score + 0.3 * sparse_score，理论最大值约 0.7。如果用 min_similarity=0.7 过滤，结果几乎全被剔除。解决方案：将阈值降到 0.3，或根据融合策略动态调整。