# Workflow Local Development ## 快速参考 ### 按键命令任务|命令 ---|--- Check pods | `kubectl get pods --context kind-kind` Restart component | `tilt trigger workflow-` View component logs | `tilt logs workflow-` ### Locally Deployed Components 这些工作流服务作为共享依赖项在本地部署：“workflow-catalog”、“workflow-executions-api”、“workflow-engine-worker”、“workflow-consumer”、“workflow-validator”、“workflows-worker”、“standalone-tasks-worker” ** ** ## 实用脚本 Execute these scripts for common operations: ### 检查 Pod 状态 bash .cursor/skills/workflow-local-dev/scripts/check-pods.sh ### Restart a Service bash .cursor/skills/workflow-local-dev/scripts/restart-service.sh <服务名称> # Example: bash .cursor/skills/workflow-local-dev/scripts/restart-service.sh catalog ### Tail Service Logs bash .cursor/skills/workflow-local-dev/scripts/tail-logs.sh <服务名称> # 示例：bash .cursor/skills/workflow-local-dev/scripts/tail-logs.shexecutions-api ### 查询数据库 bash .cursor/skills/workflow-local-dev/scripts/db-query.sh "" # 示例： bash .cursor/skills/workflow-local-dev/scripts/db-query.sh "SELECT * FROMworkflow_engine.workflows ORDER BYcreated_at DESC LIMIT 5" ** ** ## 开发工作流程主要当容器在本地失败或需要调试运行时行为时使用此流程。 1. **捕获当前状态** : `kubectl get pods --context kind-kind` 2. **重建/重启组件** : `倾斜触发工作流程-` 3. **Watch logs or health** : `kubectl logs -f --context kind-kind` 4. **检查配置或数据**（如果需要）：`bash .cursor/skills/workflow-local-dev/scripts/db-query.sh ""` ** ** ## 故障排除 ### Pod Issues kubectl get pods --context kind-kind kubectl 描述 pod --context kind-kind kubectl logs -f --context kind-kind ### Temporal Workflows 打开（临时 UI）并按工作流 ID 进行搜索。 ### Database State # Use the db-query script (runs psql inside the postgres pod) bash .cursor/skills/workflow-local-dev/scripts/db-query.sh“从工作流_engine.workflows中选择* ORDER BYcreated_at DESC LIMIT 5” ### Pulsar Messages pulsar-admin 主题统计持久://public/dap/ pulsar-client 消费持久://public/dap/ -s test-sub -n 10 ** ** ## Using Kubernetes MCP (When Available) If the Kubernetes MCP server is enabled, you can use MCP tools instead of shell commands for a more integrated experience. The MCP defaults to the `kind-kind` context. ### List Pods Tool: pods_list 参数：{ "labelSelector": "app=workflow-catalog" } Or list all pods: Tool: pods_list 参数：{} ### Get Pod Logs Tool: pods_log Arguments: { "name": "workflow-catalog-xxxxx", "namespace": "default", "tail": 100 } For previous container logs (after crash): Tool: pods_log Arguments: { "name": "workflow-catalog-xxxxx", "namespace": "default", "previous": true } ### Execute Commands in Pods (e.g., Database Queries) Tool: pods_exec Arguments: { "name": "postgres-xxxxx", "namespace": "default", "command": ["psql", "-U", "postgres", "-d", "temporal", "-c", "SELECT * FROM workflow_engine.workflows ORDER BY created_at DESC LIMIT 5"] } ### Delete Pod (Force Restart) Tool: pods_delete Arguments: { "name": "workflow-catalog-xxxxx", "namespace": "default" } ### Get Pod Details Tool: pods_get Arguments: { "name": "workflow-catalog-xxxxx", "namespace": "default" } ### List Cluster Events (Troubleshooting) Tool: events_list 参数：{“命名空间”：“默认”} > **Note** : When using MCP tools, you don't need to specify `context` as it defaults to `kind-kind`.当您希望代理直接与集群交互而不生成 shell 进程时，MCP 方法非常有用。 ** ** ## 其他资源 * For complete service URLs and infrastructure details, see [reference.md]()

工作流程本地开发

包含技能