Runtime web server monitoring
You can monitor the general health and availability of a runtime and its web server by periodically checking the runtime status endpoint:
http://<host>:<port>/_admin/status
This endpoint is available only when the web server is started. You can start the web server by deploying an API. Alternatively, you can set the following container property to force the web server to start automatically: com.boomi.container.plugin.force.start.shared_http_server=true.
This status endpoint enables you to verify that a runtime's web server is active and ready to accept incoming web service requests. The endpoint returns an HTTP status code of 200 OK if the runtime is running and all known listeners have started. Failure to connect with the endpoint or a status code other than 200 means that the runtime is not ready to accept new requests. Knowing the status of a runtime's web server can also help you balance the workload between nodes in a runtime cluster or runtime cloud more effectively.
The runtime status endpoint is not authenticated.
Status monitoring
For basic status monitoring, you can use the endpoint to verify that a runtime's web server is active and accepting incoming requests, and that the runtime itself is operational.
Load balancing
To improve workload balancing and availability in a runtime cluster or runtime cloud, you can use the status endpoint together with a property that delays runtime shutdown. Load balancers periodically check the status of each web server, but there is an inherent delay between when a runtime indicates it can no longer accept work and when the load balancer recognizes this status. During this time, the load balancer could continue sending requests to a runtime that is not able to handle them.
When monitoring the nodes in a runtime cluster or runtime cloud, call the status endpoint for individual basic runtimes from within the load balancer configuration. The status endpoint is intended for use by the load balancer itself — do not use an external web service to call the status endpoint from outside the load balancer.
Basic runtimes have the ability to enforce a delay between the time the status endpoint reports the runtime is not ready and the time that the runtime actually begins shutting down (during which the runtime can still handle incoming requests). The Runtime Pending Shutdown Delay property (com.boomi.container.pendingShutdownWarnTime) controls this delay. By default, there is no delay — the Runtime Pending Shutdown Delay default value is 0 milliseconds. If you specify a delay value greater than 0, the basic runtime continues to run during that time, but the status endpoint returns an HTTP status code of 503 Service Unavailable. This status lets the load balancer know that the runtime is not ready to accept new requests and that it should redirect requests to other nodes.
To make best use of this delay, set the Runtime Pending Shutdown Delay property to a value greater than the time it takes for your load balancer to recognize that a basic runtime is no longer available. For example, if you configure your load balancer to check the status endpoint every 5 seconds and remove a basic runtime after two failed checks, set the Runtime Pending Shutdown Delay property to a value greater than 10 seconds. With this configuration, the load balancer can balance the workload with no gaps in availability to web service clients.
