Background: The original control API (https://apisix.apache.org/zh/docs/apisix/control-api/#get-v1healthcheck) has the following issues:
* The control API may be handled by any worker process. If the handling process hasn't started the health check yet, then no data returns. But we need a global view of the whole APISIX instance, no matter which worker process responds to the control API request. https://github.com/apache/apisix/issues/5953 * No HTML response format https://github.com/apache/apisix/issues/8441 * No Prometheus metrics support Solution: When the health checker updates the status, posts the status to a global event source, and each worker process registers that event source so that each worker has a global and complete view of the health status of all resources. Note that only when one upstream is satisfied by the conditions below, its status is shown in the result list: * The upstream is configured with a health checker * The upstream has served requests in any worker process Response format: The response format is determined by the Accept header. If the header does not specify HTML explicitly, then use JSON by default. Example: [ { "name":"upstream#/apisix/routes/1", "nodes":[ { "host":"foo", "ip":"127.0.0.1", "port":9090, "status":"healthy" }, { "host":"bar", "ip":"127.0.0.4", "port":222, "status":"mostly_healthy" }, { "host":"127.0.0.2", "ip":"127.0.0.2", "port":54, "status":"unhealthy" } ] } ] Prometheus metrics example: upstream_status{name="upstream#/apisix/routes/1",host="foo",ip="127.0.0.1",port="89"} 1 upstream_status{name="upstream#/apisix/routes/demo",host="foo.bar",ip="192.168.23.2",port="334"} 0
