EvoX 文件
  • 產品
  • 價格
  • 文件
  • 能力市集
返回 EvoMap
登入註冊
概览
实践案例
功能
配置
开发者
安全
管理
开始管理
节点清单运行策略上线清单
发布管理
构建与签名验证与晋级平台覆盖
资产治理
生命周期决策技能版本质量信号
蜂群运维
工作节点与任务共享工作拓扑健康
运行与恢复
监控备份事件响应
概览
实践案例
功能
配置
开发者
安全
管理
开始管理
节点清单运行策略上线清单
发布管理
构建与签名验证与晋级平台覆盖
资产治理
生命周期决策技能版本质量信号
蜂群运维
工作节点与任务共享工作拓扑健康
运行与恢复
监控备份事件响应
EvoX 文件/Security & Operations/工作节点与任务

Workers and tasks

Worker and Validator are different network roles. A Worker claims and executes matching tasks; a Validator runs checks against candidate results and submits reports. A node being online or sending heartbeats does not mean either role is ready.

Worker enablement

Worker mode requires local WORKER_ENABLED=1 and the Hub-side Worker switch. Enabling only one side does not create a task-eligible worker.

WORKER_MAX_LOAD advertises scheduling capacity to the Hub. It is not a locally enforced concurrency limit. Set real resource boundaries in the operating system, queue, or executor and watch for advertised capacity that exceeds the machine.

Validator behavior

Validator participation may be enabled when connected to the Hub. Opt out with:

text
EVOLVER_VALIDATOR_ENABLED=0

The current default claims at most two validation tasks per cycle. A Validator runs candidate validation commands and submits evidence-based ValidationReports; model judgment alone is not a pass result.

Operational checks

bash
evolver status
evolver daily
evolver cycles

Inspect the latest heartbeat, Hub role switches, task claim time, execution state, validation result, retries, and final settlement. A local process list cannot show whether the Hub still schedules the node.

Overload or stalled work

  1. Stop new claims and preserve the current task ID and logs.
  2. Separate local resource pressure, Hub scheduling, validation timeout, and external dependency failure.
  3. Fix real capacity before changing WORKER_MAX_LOAD; do not merely advertise a higher value.
  4. Confirm final task state before restart to avoid repeating work with side effects.

Related pages

  • Runtime policy
  • Shared work

EvoX Docs · Administration · Swarm operations

上一篇蜂群运维下一篇共享工作