Dagster Deep Management Skill

Manage and monitor Dagster jobs, assets, schedules, sensors, and infrastructure health.

MANDATORY: Discovery-First Pattern

Always list jobs and check daemon health before querying specific runs.

Phase 1: Discovery

#!/bin/bash

DAGSTER_API="${DAGSTER_URL:-http://localhost:3000}/graphql"

dagster_gql() {
    curl -s -H "Content-Type: application/json" \
         ${DAGSTER_API_TOKEN:+-H "Authorization: Bearer $DAGSTER_API_TOKEN"} \
         "${DAGSTER_API}" \
         -d "{\"query\": \"$1\"}"
}

echo "=== Dagster Instance Health ==="
dagster_gql "{ instance { daemonHealth { allDaemonStatuses { daemonType healthy lastHeartbeatTime } } } }" | jq -r '
    .data.instance.daemonHealth.allDaemonStatuses[] |
    "\(.daemonType)\t\(.healthy)\t\(.lastHeartbeatTime)"
' | column -t

echo ""
echo "=== Repositories ==="
dagster_gql "{ repositoriesOrError { ... on RepositoryConnection { nodes { name location { name } } } } }" | jq -r '
    .data.repositoriesOrError.nodes[] |
    "\(.name)\t\(.location.name)"
' | column -t

echo ""
echo "=== Jobs ==="
dagster_gql "{ repositoriesOrError { ... on RepositoryConnection { nodes { name pipelines { name } } } } }" | jq -r '
    .data.repositoriesOrError.nodes[] |
    .name as $repo | .pipelines[] |
    "\($repo)\t\(.name)"
' | column -t | head -30

echo ""
echo "=== Schedules ==="
dagster_gql "{ schedulesOrError { ... on Schedules { results { name scheduleState { status } cronSchedule } } } }" | jq -r '
    .data.schedulesOrError.results[] |
    "\(.name)\t\(.scheduleState.status)\t\(.cronSchedule)"
' | column -t | head -20

Phase 2: Analysis

#!/bin/bash

echo "=== Failed Runs (recent) ==="
dagster_gql "{ runsOrError(filter: { statuses: [FAILURE] }, limit: 15) { ... on Runs { results { runId pipelineName status startTime endTime } } } }" | jq -r '
    .data.runsOrError.results[] |
    "\(.runId[:8])\t\(.pipelineName)\t\(.status)\t\(.startTime)"
' | column -t

echo ""
echo "=== Sensors ==="
dagster_gql "{ sensorsOrError { ... on Sensors { results { name sensorState { status } } } } }" | jq -r '
    .data.sensorsOrError.results[] |
    "\(.name)\t\(.sensorState.status)"
' | column -t

echo ""
echo "=== Asset Health ==="
dagster_gql "{ assetsOrError { ... on AssetConnection { nodes { key { path } assetMaterializations(limit: 1) { timestamp } } } } }" | jq -r '
    .data.assetsOrError.nodes[] |
    "\(.key.path | join("/"))\tLast: \(.assetMaterializations[0].timestamp // "never")"
' | column -t | head -20

echo ""
echo "=== Run Queue ==="
dagster_gql "{ runsOrError(filter: { statuses: [QUEUED, STARTING] }, limit: 10) { ... on Runs { results { runId pipelineName status } } } }" | jq -r '
    .data.runsOrError.results[] |
    "\(.runId[:8])\t\(.pipelineName)\t\(.status)"
' | column -t

Output Rules

• TOKEN EFFICIENCY: Target <=50 lines per output
• Use status filters on GraphQL queries
• Never dump full run logs or asset metadata -- extract key status fields

Output Format

Present results as a structured report:

Managing Dagster Deep Report
════════════════════════════
Resources discovered: [count]

Resource       Status    Key Metric    Issues
──────────────────────────────────────────────
[name]         [ok/warn] [value]       [findings]

Summary: [total] resources | [ok] healthy | [warn] warnings | [crit] critical
Action Items: [list of prioritized findings]

Target ≤50 lines of output. Use tables for multi-resource comparisons.

Anti-Hallucination Rules

1. NEVER assume resource names — always discover via CLI/API in Phase 1 before referencing in Phase 2.
2. NEVER fabricate metric names or dimensions — verify against the service documentation or --help output.
3. NEVER mix CLI commands between service versions — confirm which version/API you are targeting.
4. ALWAYS use the discovery → verify → analyze chain — every resource referenced must have been discovered first.
5. ALWAYS handle empty results gracefully — an empty response is valid data, not an error to retry.

Counter-Rationalizations

Shortcut	Counter	Why
"I'll skip discovery and check known resources"	Always run Phase 1 discovery first	Resource names change, new resources appear — assumed names cause errors
"The user only asked for a quick check"	Follow the full discovery → analysis flow	Quick checks miss critical issues; structured analysis catches silent failures
"Default configuration is probably fine"	Audit configuration explicitly	Defaults often leave logging, security, and optimization features disabled
"Metrics aren't needed for this"	Always check relevant metrics when available	API/CLI responses show current state; metrics reveal trends and intermittent issues
"I don't have access to that"	Try the command and report the actual error	Assumed permission failures prevent useful investigation; actual errors are informative

Common Pitfalls

• Daemon health: Scheduler and sensor daemons must be running -- stale heartbeats mean no scheduling
• Run coordinator: Run queue limits control concurrency -- queued runs wait for slots
• Code location errors: Import errors prevent jobs from loading -- check code location health
• Partition backfills: Large backfills can overwhelm the run coordinator
• Asset freshness: Stale assets may indicate upstream failures -- trace the dependency graph
• Sensor cursor: Sensor cursors track state -- resetting a cursor reprocesses events
• Resource config: Missing or invalid resource configurations cause immediate run failures
• IO managers: IO manager misconfiguration causes silent data loss between ops