When creating a Dataproc cluster, you can specify initialization actions in executables and/or scripts that Dataproc will run on all nodes in your Dataproc cluster immediately after the cluster is set up. Initialization actions often set up job dependencies, such as installing Python packages, so that jobs can be submitted to the cluster without having to install dependencies when the jobs are run.
Initialization actions must be stored in a Cloud Storage bucket and can be passed as a parameter to the gcloud command or the clusters.create API when creating a Dataproc cluster. For example, to specify an initialization action when creating a cluster with the gcloud command, you can run:
gcloud dataproc clusters create <CLUSTER_NAME> \
[--initialization-actions [GCS_URI,...]] \
[--initialization-action-timeout TIMEOUT]During development, you can create a Dataproc cluster using Dataproc-provided
regional initialization
actions buckets (for example goog-dataproc-initialization-actions-us-east1):
REGION=<region>
CLUSTER=<cluster_name>
gcloud dataproc clusters create ${CLUSTER} \
--region ${REGION} \
--initialization-actions gs://goog-dataproc-initialization-actions-${REGION}/presto/presto.shBUCKET=<your_init_actions_bucket>
CLUSTER=<cluster_name>
gsutil cp presto/presto.sh gs://${BUCKET}/
gcloud dataproc clusters create ${CLUSTER} --initialization-actions gs://${BUCKET}/presto.shYou can decide when to sync your copy of the initialization action with any changes to the initialization action that occur in the GitHub repository. Doing this is also useful if you want to modify initialization actions to meet your needs.
These samples are provided to show how various packages and components can be installed on Dataproc clusters. You should understand how these samples work before running them on your clusters. The initialization actions provided in this repository are provided without support and you use them at your own risk.
This repository currently offers the following actions for use with Dataproc clusters.
- Install additional Apache Hadoop ecosystem components
- Improve data science and interactive experiences
- Configure the environment
- Configure a nice shell environment
- To switch to Python 3, use the conda initialization action
- Connect to Google Cloud Platform services
- Install alternate versions of the Cloud Storage and BigQuery connectors. Specific versions of these connectors come pre-installed on Cloud Dataproc clusters.
- Share a Cloud SQL Hive Metastore, or simply read/write data from Cloud SQL.
- Set up monitoring
Previously, this repo provided init actions for the following, which have since been removed because equivalent functionality is now provided directly by Dataproc:
- Apache Tez: This is now pre-installed in all current Dataproc image versions.
- Datalab: Datalab has been replaced by Vertex AI Workbench, which integrates with Dataproc.
- Jupyter: This has been replaced with the Jupyter Optional Component.
Single Node clusters have dataproc-role set to Master and dataproc-worker-count set to 0. Most of the initialization actions in this repository should work out of the box because they run only on the master. Examples include notebooks, such as Apache Zeppelin, and libraries, such as Apache Tez. Actions that run on all nodes of the cluster, such as cloud-sql-proxy, also work out of the box.
Some initialization actions are known not to work on Single Node clusters. All of these expect to have daemons on multiple nodes.
- Apache Drill
- Apache Flink
- Apache Kafka
- Apache Zookeeper
Feel free to send pull requests or file issues if you have a good use case for running one of these actions on a Single Node cluster.
Dataproc sets special metadata values for the instances that run in your cluster. You can use these values to customize the behavior of initialization actions, for example:
ROLE=$(/usr/share/google/get_metadata_value attributes/dataproc-role)
if [[ "${ROLE}" == 'Master' ]]; then
... master specific actions ...
else
... worker specific actions ...
fiYou can also use the ‑‑metadata flag of the gcloud dataproc clusters create command to provide your own
custom metadata:
gcloud dataproc clusters create cluster-name \
--initialization-actions ... \
--metadata name1=value1,name2=value2,... \
... other flags ...For more information, review the Dataproc documentation. You can also pose questions to the Stack Overflow community with the tag google-cloud-dataproc.
See our other Google Cloud Platform github
repos for sample applications and
scaffolding for other frameworks and use cases.
Subscribe to cloud-dataproc-discuss@google.com for announcements and discussion.
- See CONTRIBUTING.md
- See LICENSE
- You might see an error message similar to the following when upgrading the agent, installing the agent, or running apt-get update on Debian/Ubuntu Linux:
E: Repository 'https://packages.cloud.google.com/apt google-cloud-monitoring-buster-all InRelease' changed its 'Origin' value from 'google-cloud-monitoring-buster' to 'namespaces/cloud-ops-agents-artifacts/repositories/google-cloud-monitoring-buster-all'
E: Repository 'https://packages.cloud.google.com/apt google-cloud-monitoring-buster-all InRelease' changed its 'Label' value from 'google-cloud-monitoring-buster' to 'namespaces/cloud-ops-agents-artifacts/repositories/google-cloud-monitoring-buster-all'
This message indicates that the package repository cache may have diverged from its source. To resolve this, run the following command:
apt-get --allow-releaseinfo-change update
Then, run the upgrade or install again.