DEADLINE_EXCEEDED error when indexing codebase - gRPC timeout during collection creation

[aud1ence]

Bug Description

When trying to index a codebase using the index_codebase tool, it consistently fails with a gRPC DEADLINE_EXCEEDED error during collection creation validation.

Error Message

Error validating collection creation: 4 DEADLINE_EXCEEDED: Deadline exceeded after 15.001s, metadata filters: 0.001s, remote_addr=34.111.198.99:443

Environment

• MCP Package: @zilliz/claude-context-mcp@latest
• Embedding Provider: VoyageAI (voyage-code-3)
• Milvus Address: Zilliz serverless cluster (*.serverless.gcp-us-west1.cloud.zilliz.com)
• OS: macOS Darwin 24.6.0
• Claude Code: CLI

Steps to Reproduce

1. Configure claude-context MCP with a Zilliz serverless cluster
2. Call index_codebase with an absolute path to a local directory
3. Error is thrown immediately after 15 seconds

Network Diagnostics

The Zilliz cluster endpoint is reachable:

• ✅ DNS resolves correctly
• ✅ TCP connection to port 443 succeeds
• ✅ TLS handshake completes successfully (TLSv1.3)
• ✅ Ping responds with ~22ms latency, 0% packet loss

The issue appears to be at the gRPC layer — the HTTP/HTTPS connection works fine, but the gRPC call times out during collection creation validation.

Expected Behavior

The codebase should be indexed successfully.

Actual Behavior

The tool times out after 15 seconds with DEADLINE_EXCEEDED on every attempt.

Possible Cause

The default gRPC deadline of 15 seconds may be too short, especially when:

• The Zilliz serverless cluster is warming up / resuming from a paused state
• There is high latency on the gRPC metadata/collection API

Suggested Fix

• Increase the gRPC deadline timeout for collection creation validation (e.g., 30–60 seconds)
• Add retry logic with exponential backoff for transient timeout errors
• Provide a clearer error message indicating the cluster may be paused/warming up

[aud1ence]

Related Issues

Found several existing issues reporting the same or similar problem:

Directly related (same error)

• DEADLINE_EXCEEDED on checkCollectionLimit() - gRPC timeout on collection creation (Free tier, macOS) #289 — DEADLINE_EXCEEDED on checkCollectionLimit() - gRPC timeout on collection creation (Free tier, macOS) (opened 1 day ago)

  • Identical error on the same remote addr 34.111.198.99:443
  • Root cause traced to checkCollectionLimit() in milvus-vectordb.js:607 using a dummy create/drop collection as a pre-flight check with no configurable timeout
  • Suggests replacing with listCollections() + count to avoid heavy gRPC call

• Indexing error : Error validating collection creation #215 — Indexing error : Error validating collection creation (open)

  • Same error: Error validating collection creation

Suggested fix (from #289)

The root cause is checkCollectionLimit() creating a dummy collection to test limits, which triggers a gRPC call that exceeds the hardcoded 15s deadline. Fix options:

1. Configurable timeout via env var (e.g. MILVUS_TIMEOUT_MS)
2. Replace dummy collection check with listCollections() + count (less expensive)
3. REST API fallback — MilvusRestfulVectorDatabase exists but isn't wired as fallback