How to create a Dataproc cluster with time to live using Python SDK

Question

I try to create a Dataproc cluster which has a time to live of 1 day using python SDK. For this purpose, v1beta2 of the Dataproc API introduces the LifecycleConfig object which is child of the ClusterConfig object.

I use this object in the JSON file which I pass to the create_cluster method. To set the particular TTL, I use the field auto_delete_ttl which shall have the value 86,400 seconds (one day).

The documentation of Google Protocol Buffers is rather specific about how to represent a duration in the JSON file: Durations shall be represented as string with suffix s for seconds and there shall be 0,3,6 or 9 fractional seconds:

However, if I pass the duration using this format, I get the error:

Parameter to MergeFrom() must be instance of same class: expected google.protobuf.Duration got str

This is how I create the cluster:

from google.cloud import dataproc_v1beta2
project = "your_project_id"
region = "europe-west4"
cluster = "" #see below for cluster JSON file
client = dataproc_v1beta2.ClusterControllerClient(client_options={
    'api_endpoint': '{}-dataproc.googleapis.com:443'.format(region)
})

# Create the cluster
operation = client.create_cluster(project, region, cluster)

The variable cluster holds the JSON object describing the desired cluster:

{
  "cluster_name":"my_cluster",
  "config":{
     "config_bucket":"my_conf_bucket",
     "gce_cluster_config":{
        "zone_uri":"europe-west4-a",
        "metadata":{
           "PIP_PACKAGES":"google-cloud-storage google-cloud-bigquery"
        },
        "subnetwork_uri":"my subnet",
        "service_account_scopes":[
           "https://www.googleapis.com/auth/cloud-platform"
        ],
        "tags":[
           "some tags"
        ]
     },
     "master_config":{
        "num_instances":1,
        "machine_type_uri":"n1-highmem-4",
        "disk_config":{
           "boot_disk_type":"pd-standard",
           "boot_disk_size_gb":200,
           "num_local_ssds":0
        },
        "accelerators":[

        ]
     },
     "software_config":{
        "image_version":"1.4-debian9",
        "properties":{
           "dataproc:dataproc.allow.zero.workers":"true",
           "yarn:yarn.log-aggregation-enable":"true",
           "dataproc:dataproc.logging.stackdriver.job.driver.enable":"true",
           "dataproc:dataproc.logging.stackdriver.enable":"true",
           "dataproc:jobs.file-backed-output.enable":"true"
        },
        "optional_components":[

        ]
     },
     "lifecycle_config":{
        "auto_delete_ttl":"86400s"
     },
     "initialization_actions":[
        {
           "executable_file":"gs://some-init-script"
        }
     ]
  },
  "project_id":"project_id"
  }

Package versions I am using:

• google-cloud-dataproc: 0.6.1
• protobuf: 3.11.3
• googleapis-common-protos: 1.6.0

Am I doing something wrong here, is it an issue with wrong package versions or is it even a bug?

Answer 1

You should use 100s format for a duration type when you construct protobuf in a text format (i.e. json, etc), but you are using a Python object to construct API request body, that's why you need to create a Duration object instead of a string:

duration_message.FromSeconds(86400)