Debug

Goals

• Find why a run is stuck, retrying, or failing.
• Correlate Linear issue identity to an OpenHands session quickly.
• Read the right logs in the right order to isolate root cause.

Log Sources

• Primary runtime log: log/symphony.log
  • Includes orchestrator, agent runner, and OpenHands runtime lifecycle logs.
• Rotated runtime logs: log/symphony.log*
  • Check these when the relevant run is older.

Correlation Keys

• issue_identifier: human ticket key (example: MT-625)
• issue_id: Linear UUID (stable internal ID)
• conversation_id: OpenHands conversation UUID
• session_id: OpenHands thread-turn pair (<thread_id>-<turn_id>)

Use these fields as join keys during debugging.

Quick Triage (Stuck Run)

1. Confirm scheduler/worker symptoms for the ticket.
2. Find recent lines for the ticket (issue_identifier first).
3. Extract conversation_id or session_id from matching lines.
4. Trace that ID across start, stream, completion/failure, and stall handling logs.
5. Decide class of failure: timeout/stall, runtime startup failure, turn failure, or orchestrator retry loop.

Commands

# 1) Narrow by ticket key (fastest entry point)
rg -n "issue_identifier=MT-625" log/symphony.log*

# 2) If needed, narrow by Linear UUID
rg -n "issue_id=<linear-uuid>" log/symphony.log*

# 3) Pull conversation IDs seen for that ticket
rg -o "conversation_id=[^ ;]+" log/symphony.log* | sort -u

# 4) Trace one conversation end-to-end
rg -n "conversation_id=<uuid>" log/symphony.log*

# 5) Focus on stuck/retry signals
rg -n "Issue stalled|scheduling retry|turn_timeout|turn_failed|session failed|session ended with error" log/symphony.log*

Investigation Flow

1. Locate the ticket slice:
   • Search by issue_identifier=<KEY>.
   • If noise is high, add issue_id=<UUID>.
2. Establish timeline:
   • Identify first session start for the issue.
   • Follow with completion, error, or worker exit lines.
3. Classify the problem:
   • Stall loop: Issue stalled ... restarting with backoff.
   • Runtime startup failure: session creation/attach failures.
   • Turn execution failure: turn_failed, turn_cancelled, turn_timeout.
   • Worker crash: Agent task exited ... reason=....
4. Validate scope:
   • Check whether failures are isolated to one issue/session or repeating across multiple tickets.
5. Capture evidence:
   • Save key log lines with timestamps, issue_identifier, issue_id, and conversation_id.
   • Record probable root cause and the exact failing stage.

Reading Session Logs

OpenSymphony session diagnostics are emitted into log/symphony.log and keyed by conversation_id. Read them as a lifecycle:

1. Session start: conversation_id=... created/attached
2. Session stream/lifecycle events for the same conversation_id
3. Terminal event:
   • Session completed successfully, or
   • Session ended with error, or
   • Issue stalled ... restarting with backoff

For one specific session investigation, keep the trace narrow:

1. Capture one conversation_id for the ticket.
2. Build a timestamped slice for only that conversation:
   • rg -n "conversation_id=<uuid>" log/symphony.log*
3. Mark the exact failing stage:
   • Startup failure before stream events (connection/attach errors).
   • Turn/runtime failure after stream events (turn_* / error events).
   • Stall recovery (Issue stalled ... restarting with backoff).
4. Pair findings with issue_identifier and issue_id from nearby lines to confirm you are not mixing concurrent retries.

Always pair session findings with issue_identifier/issue_id to avoid mixing concurrent runs.

Notes

• Prefer rg over grep for speed on large logs.
• Check rotated logs (log/symphony.log*) before concluding data is missing.
• If required context fields are missing in new log statements, align with logging conventions.