Package org.apache.flink.runtime.rest.handler.job.checkpoints

Class CheckpointHandlers

  • public class CheckpointHandlers
extends Object
    HTTP handlers for asynchronous triggering of checkpoints.

    Drawing checkpoints is a potentially long-running operation. To avoid blocking HTTP connections, checkpoints must be drawn in two steps. First, an HTTP request is issued to trigger the checkpoint asynchronously. The request will be assigned a TriggerId, which is returned in the response body. Next, the returned TriggerId should be used to poll the status of the checkpoint until it is finished.

    A checkpoint is triggered by sending an HTTP POST request to /jobs/:jobid/checkpoints. The HTTP request may contain a JSON body to specify a customized TriggerId and a CheckpointType, e.g., 

     { "triggerId": "7d273f5a62eb4730b9dea8e833733c1e", "checkpointType": "FULL" }

    If the body is omitted, or the field checkpointType is null, the default checkpointType of CheckpointType.CONFIGURED will be used. As written above, the response will contain a request id, e.g., 

     { "request-id": "7d273f5a62eb4730b9dea8e833733c1e" }

    To poll for the status of an ongoing checkpoint, an HTTP GET request is issued to /jobs/:jobid/checkpoints/:checkpointtriggerid. If the specified checkpoint is still ongoing, the response will be 

     {
     "status": {
         "id": "IN_PROGRESS"
     }
 }

    If the specified checkpoints has completed, the status id will transition to COMPLETED, and the response will additionally contain information about the savepoint, such as the location: 

     {
     "status": {
         "id": "COMPLETED"
     },
     "operation": {
         "checkpointId": "123"
     }
 }

    • Constructor Detail

      • CheckpointHandlers

        public CheckpointHandlers()