deployment-operations
Use when a coding agent needs to configure or troubleshoot Dagster OSS production operations: DAGSTER_HOME, dagster.yaml, instance storage, webserver and daemon services, run launchers/coordinators/executors, retries, run monitoring, Docker/Kubernetes/ECS deployment shape, or code-location deployment failures.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- deployment-operations
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Use when a coding agent needs to configure or troubleshoot Dagster OSS production operations: DAGSTER_HOME, dagster.yaml, instance storage, webserver and daemon services, run launchers/coordinators/executors, retries, run monitoring, Docker/Kubernetes/ECS deployment shape, or code-location deployment failures.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/troubleshooting.md9.9 KB
- 参考文档references/workflows.md10.8 KB
- 脚本scripts/check_dagster_instance_config.py6.5 KB
03 SKILL.md 正文
Deployment Operations
Use this sub-skill for Dagster OSS deployment and operations work after the user has working Dagster definitions and wants persistent services, durable instance state, queued or remote run execution, schedule/sensor automation, monitoring, retries, or infrastructure-specific deployment guidance.
Route First
- Use this sub-skill for
DAGSTER_HOME,dagster.yaml,dagster-webserver,dagster-daemon, instance storage, run queue/concurrency, retries/timeouts, run monitoring, and run launcher operations. - Use this sub-skill for Docker, Kubernetes, and ECS deployment checklists at the concept/configuration level, including image, workspace, service, and secret concerns.
- Route local command usage,
dagster dev, definitions validation, and one-off asset/job execution to../cli-local-development/SKILL.mdif that sub-skill exists. - Route schedule/sensor definition authoring to
../automation-schedules-sensors/SKILL.mdif that sub-skill exists; use this sub-skill when the issue is daemon/runtime operation. - Route resource/run config authoring and environment-variable config schemas to
../configuration-resources/SKILL.mdif that sub-skill exists. - Do not cover Dagster Plus account setup, Dagster Cloud CLI deployment, UI TypeScript internals, or Helm template maintenance here.
Start Here
- Confirm the target is Dagster OSS and identify the deployment shape: local service, Docker Compose, Kubernetes, ECS, or custom process supervisor.
- Establish one shared
DAGSTER_HOMEfor the webserver, daemon, and run workers; verify thatdagster.yamllives there and is readable by every service. - Run
python scripts/check_dagster_instance_config.py --dagster-home <path>from this sub-skill to check for a safe baseline without starting services. - Decide whether the instance needs durable external storage. SQLite is acceptable for local/single-node evaluation; production multi-service deployments usually need shared database-backed storage and shared compute logs.
- Confirm the long-running services are represented: one or more webservers, exactly one daemon process, and one code-location server per code location.
- Configure execution flow deliberately: run coordinator queues/submits runs, daemon dequeues if queued, run launcher starts the run worker, and the job executor controls per-step execution inside the worker.
References
- Deployment workflows for service topology,
dagster.yamlpatterns, run queue/concurrency, retries, monitoring, daemon/webserver operations, and Docker/Kubernetes/ECS deployment checklists. - Troubleshooting for import/install issues, optional dependency gaps, config misuse, daemon heartbeat failures, storage mismatches, code-location failures, image pull failures, stuck runs, and concurrency deadlocks.
- Instance config checker for a safe
DAGSTER_HOMEanddagster.yamlinspection helper that does not start services or contact infrastructure.
Safety Notes
- Do not start or restart production services, run migrations, wipe storage, edit secrets, push images, or apply cluster/ECS changes without explicit user approval.
- Treat
dagster instance migrate, direct database cleanup, Kubernetes/ECS deploy commands, and Docker Composeup/downas mutating operations. - Keep secrets in environment variables or platform secret stores; do not write credentials directly into public
dagster.yaml, workspace files, Dockerfiles, manifests, or generated skill content.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).