Skip to content

Latest commit

 

History

History
116 lines (101 loc) · 4.29 KB

dataproc.md

File metadata and controls

116 lines (101 loc) · 4.29 KB
Raw

Dataproc

  • Managed Hadoop (Spark, SparkML, Hive, Pig, etc…)

  • Automated cluster management, resizing

  • Code/Query only

  • Job management screen in the console

  • Think in terms of a ‘job specific resource’ – for each job, create a cluster and then delete it.

  • Used if migrating existing on-premise Hadoop or Spark infrastructure to GCP without redevelopment effort.

  • Can sale even when jobs are running.

  • Use Dataflow for streaming instead. This is better for batch.

    alt text

  • Connecting to Web Interface of Dataproc Cluster

    • Allow necessary web ports access via firewall rules, and limit access to your network.
      • Tcp:8088 (Cluster Manager)
        • :8088
      • Tcp:50070 (Connect to HDFS name node)
        • :50070
    • OR SOCKS proxy (routes through an SSH tunnel for secure access)
      • gcloud compute ssh [master_node_name]

Cost

  • Standard Compute Engine machine type pricing + managed Dataproc premium.
  • Premium = $0.01 per vCPU core/hour
  • Billed by the second, with a minimum of 1 minute.

Storage

  • Can use on disk (HDFS) or GCS

HDFS

  • Split up on the cluster, but requires cluster to be up.

GCS

  • Allows for the use of preemptible machines that can reduce costs significantly.
    • DO not need to configure startup and shutdown scripts to gracefully handle shutdown, Dataproc already handles this.
    • Cluster MUST have at least 2 standard worker nodes however.
  • Separate cluster and storage.
  • Cloud Storage Connector
    • Allows you to run Hadoop or Spark jobs directly on GCS.
    • Quick startup
      • In HDFS, a MapReduce job can’t start until the NameNode is out of safe mode.
      • With GCS, can start job as soon as the task nodes start, leading to significant cost savings over time.

  • Cluster Machine Types

    • Build using Compute Engine VM instances
    • Cluster – need at least 1 master and 2 workers

  • High Availability Mode

    • 3 masters rather than 1
    • 3 masters run in an Apache Zookeeper cluster for automatic failover.

  • Restartable Jobs

    • Jobs do NOT restart on failure (default)
    • Can change this – useful for long running and streaming jobs (ex. Spark Streaming)
    • Mitigates out-of-memory errors, unscheduled reboots

Updating Clusters

  • Can only change # workers/preemptible VM’s/labels/toggle graceful decommission.
    • Graceful Decommissioning
      • Finish work in progress on a worker before it is removed from Dataproc cluster.
      • Incorporates graceful YARN decommissioning.
      • May fail for preemptible workers.
        • Can forcefully decommission preemptible workers at any time.
      • Will always work with primary workers.
      • gcloud dataproc clusters update –gracefult-decommission-timeout
        • Default to “0s” – forceful decommissioning.
        • Need to provide a time.
        • Max value 1 day.
    • Automatically reshards data for you.

Migrating and Optimizing for GCP Best Practices

  • Move data first
    • Generally to GCS buckets.
    • Possible exceptions
      • Apache HBase data to BigTable
      • Apache Impala to BigQuery
      • Can still choose to move to GCS if BigTable/BQ feature not needed.
  • Small scale experimentation
    • Use a subset of data to test.
  • Think of it in terms of ephemeral clusters.
  • Use GCP tools to optimize and save costs.

Converting from HDFS to GCS

  • Copy data to GCS
    • Install connector or copy manually
  • Update file prefix in scripts
    • Hdfs:// to gs://
  • Use Dataproc, and run against/output to GCS

Connectors

  • BQ/BigTable (copies data to GCS) /CloudStorage

Optional Components

  • Anaconda, Druid
  • Hive WebHCat
  • Jupyter
  • Kerberos
  • Presto
  • Zeppelin
  • Zookeeper
  • How to configure Hadoop to use all cores?
    • Think spark executor cores
  • How to handle out of memory errors?
    • Hint - Executor memory
  • How to install other components?
    • Hint – Initialization actions

IAM

  • Project level only (primitive and predefined roles)
  • Cloud Dataproc Editor, Viewer, and Worker
    • Editor – Full access to create/edit/delete clusters/jobs/workflows
    • Viewer – View access only
    • Worker – Assigned to service accounts
      • Read/write GCS, write to Cloud Logging