Skip to content

[cli-tools-test] logs tool times out with context deadline exceeded when no workflow_name filter is applied #46990

Description

@github-actions

Problem Description

The logs tool from the agentic-workflows MCP server consistently times out with context deadline exceeded when called without a workflow_name filter, even when count and start_date limits are applied.

Tool

  • Tool: logs
  • Command: agenticworkflows logs --start_date "-1d" --count 3 --max_tokens 3000

Steps to Reproduce

  1. Call the logs tool without workflow_name filter:
    agenticworkflows logs --start_date "-1d" --count 3 --max_tokens 3000
    
  2. Observe timeout after exactly 60 seconds.

Also reproduces with --start_date "-7d".

Expected Behavior

The count: 3 parameter should limit the number of runs fetched and return quickly, regardless of whether a workflow name filter is supplied. The tool should respect pagination/count limits without downloading all runs first.

Actual Behavior

The tool hangs for exactly 60 seconds then exits with exit code 1:

[info] [agenticworkflows] MCP tools/call: status=200, elapsed=60029ms
Error [0]: context deadline exceeded
[error] [agenticworkflows] Tool call error: Error [0]: context deadline exceeded

Contrast: Filtered Call Works Fine

When workflow_name is provided, the tool completes in ~1-9 seconds:

agenticworkflows logs --start_date "-1d" --count 3 --max_tokens 3000 --workflow_name "daily-cli-tools-tester"
# Completes in ~1.5s ✅

Environment

  • Repository: github/gh-aw
  • Run ID: 29805908218
  • Date: 2026-07-21
  • gh-aw version: 1.0.71 (engine: copilot, model: claude-sonnet-4.6)
  • Firewall version: v0.27.37

Impact

  • Severity: High — the primary use case for logs (browsing recent runs across all workflows) is completely broken
  • Frequency: Always (100% reproducible)
  • Workaround: Always supply --workflow_name filter, but this defeats the purpose of broad log discovery

Root Cause Hypothesis

The 60-second timeout matches the MCP client default deadline. The server appears to be fetching/scanning all workflow runs before applying the count limit server-side, rather than stopping early once the limit is reached.

Generated by 🧪 Daily Cli Tools Tester · 41.8 AIC · ⌖ 7 AIC · ⊞ 9.3K ·

  • expires on Jul 27, 2026, 10:15 PM UTC-08:00

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions