{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "## Setup\n", "\n", "1. Create a Spark session\n", "2. Add the iceberg-runtime Jar" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "Waiting for a Spark session to start..." ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/html": [ "\n", "\n" ], "text/plain": [ "\n", "Spark application_1506444763486_1280232:\n", "* http://hadoop-historyserver:20888/proxy/application_1506444763486_1280232\n", "* http://hadoop-resourcemanager:8088/cluster/app/application_1506444763486_1280232\n", "\n", "Local logs:\n", "* /data/tmp/genie/jobs/20180110_171610.028692.log\n", "* Also available using %tail_log\n" ] }, "execution_count": 1, "metadata": {}, "output_type": "execute_result" } ], "source": [ "spark" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Starting download from file:///home/user/iceberg-runtime-0.1.3.jar\n", "Finished download of iceberg-runtime-0.1.3.jar\n" ] } ], "source": [ "%AddJar file:///home/user/iceberg-runtime-0.1.3.jar" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Drop and create a table in HDFS\n", "\n", "[Spark Schema Helpers](https://netflix.github.io/iceberg/current/javadoc/index.html?com/netflix/iceberg/spark/SparkSchemaUtil.html)" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "path = hdfs:/tmp/tables/job_metrics_tmp\n" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "table {\n", " 1: event_utc_ms: optional long\n", " 2: hostname: optional string\n", " 3: jobflow: optional string\n", " 4: job_name: optional string\n", " 5: application_type: optional string\n", " 6: record_id: optional string\n", " 7: record_type: optional string\n", " 8: user: optional string\n", " 9: submit_time: optional long\n", " 10: start_time: optional long\n", " 11: finish_time: optional long\n", " 12: run_host: optional string\n", " 13: submit_host: optional string\n", " 14: status: optional string\n", " 15: cluster_id: optional string\n", " 16: cluster_name: optional string\n", " 17: queue: optional string\n", " 18: genie_job_name: optional string\n", " 19: genie_job_id: optional string\n", " 20: job_uuid: optional string\n", " 21: counters: optional string\n", " 22: properties: optional string\n", " 23: dateint: optional int\n", " 24: hour: optional int\n", " 25: batchid: optional string\n", "}" ] }, "execution_count": 3, "metadata": {}, "output_type": "execute_result" } ], "source": [ "import org.apache.hadoop.fs.Path\n", "import org.apache.iceberg.hadoop.HadoopTables\n", "import org.apache.iceberg.spark.SparkSchemaUtil\n", "\n", "val path = \"hdfs:/tmp/tables/job_metrics_tmp\"\n", "\n", "{ // use a block to avoid values (conf, etc.) getting caught in closures\n", "\n", " // remove the temp table if it already exists\n", " val conf = spark.sessionState.newHadoopConf()\n", " val fs = new Path(path).getFileSystem(conf)\n", " fs.delete(new Path(path), true /* recursive */ )\n", "\n", " // create the temp table using Spark utils to create a schema and partition spec\n", " val tables = new HadoopTables(conf)\n", " val schema = SparkSchemaUtil.schemaForTable(spark, \"default.job_metrics\")\n", " val spec = SparkSchemaUtil.specForTable(spark, \"default.job_metrics\")\n", "\n", " tables.create(schema, spec, path)\n", "\n", " // show the schema\n", " tables.load(path).schema\n", "}\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Load table partitions as a DataFrame" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", "\n", "\n", "\n", "\n", "\n", "\n", "\n", "\n", "\n", "\n", "\n", "
partitionuriformat
{dateint -> 20170316, hour -> 0, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=0/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 1, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=1/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 2, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=2/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 3, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=3/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 4, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=4/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 5, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=5/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 6, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=6/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 7, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=7/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 8, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=8/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
{dateint -> 20170316, hour -> 9, batchid -> merged_1}s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=9/batchid=merged_1org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe
" ], "text/plain": [ "+-------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------+\n", "| partition | uri | format |\n", "+-------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------+\n", "| {dateint -> 20170316, hour -> 0, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=0/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 1, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=1/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 2, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=2/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 3, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=3/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 4, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=4/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 5, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=5/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 6, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=6/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 7, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=7/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 8, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=8/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "| {dateint -> 20170316, hour -> 9, batchid -> merged_1} | s3n://bucket/hive/warehouse/job_metrics/dateint=20170316/hour=9/batchid=merged_1 | org.apache.hadoop.hive.ql.io.parquet.serde.ParquetHiveSerDe |\n", "+-------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------+-------------------------------------------------------------+" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "partitions = [partition: map, uri: string ... 1 more field]\n" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "[partition: map, uri: string ... 1 more field]" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "import org.apache.iceberg.spark.SparkTableUtil\n", "\n", "// get a data frame with the table's partitions\n", "val partitions = SparkTableUtil.partitionDF(spark, \"default.job_metrics\")\n", " .filter($\"format\".contains(\"parquet\") || $\"format\".contains(\"avro\"))\n", "\n", "display(partitions.limit(10))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# List files, load metrics, and append to the table\n", "\n", "* [Table API](https://netflix.github.io/iceberg/current/javadoc/index.html?com/netflix/iceberg/Table.html)\n", "* [Append API](https://netflix.github.io/iceberg/current/javadoc/index.html?com/netflix/iceberg/AppendFiles.html)" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[Stage 3:====================================================> (9 + 1) / 10]" ] }, { "data": { "text/plain": [ "0" ] }, "execution_count": 5, "metadata": {}, "output_type": "execute_result" } ], "source": [ "import org.apache.iceberg.hadoop.HadoopTables\n", "import org.apache.hadoop.conf.Configuration\n", "\n", "partitions.repartition(100).flatMap { row =>\n", "\n", " // list the partition and read Parquet footers to get metrics\n", " SparkTableUtil.listPartition(row.getMap[String, String](0).toMap, row.getString(1), row.getString(2))\n", "\n", "}.repartition(10) // avoid lots of manifests that would be merged later\n", " .mapPartitions { files =>\n", "\n", " // open the table and append the files from this partition\n", " val tables = new HadoopTables(new Configuration())\n", " val table = tables.load(\"hdfs:/tmp/tables/job_metrics_tmp\")\n", "\n", " // fast appends will create a manifest for the new files\n", " val append = table.newFastAppend\n", "\n", " files.foreach { file =>\n", " append.appendFile(file.toDataFile(table.spec))\n", " }\n", "\n", " // commit the new files\n", " append.commit()\n", "\n", " Seq.empty[String].iterator\n", "\n", "}.count\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# Inspect the results\n", "\n", "[Snapshot API](https://netflix.github.io/iceberg/current/javadoc/index.html?com/netflix/iceberg/Snapshot.html)" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "tables = org.apache.iceberg.hadoop.HadoopTables@1782cb95\n", "table = hdfs:/tmp/tables/job_metrics_tmp\n" ] }, "metadata": {}, "output_type": "display_data" }, { "data": { "text/plain": [ "BaseSnapshot{id=1515605124481, timestamp_ms=1515605127199, manifests=[hdfs:/tmp/tables/job_metrics_tmp/metadata/695d8ab7-961c-4cef-94d7-367db5d8f7de-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/725154b3-92bd-4d00-9420-34a2866f2876-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/266e6040-d8ff-4713-92cb-0d806c7a3baf-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/3b0e9c88-03b0-4032-bf70-f9af43e00034-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/0747127e-895e-492e-b07e-a54627ee5534-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/db055992-1bf1-4fe7-a851-1eff0a05af55-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/1d5b7cb9-85bd-4088-ad26-a4e9562ad181-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/36db4143-8720-4060-9a8d-d17fa7dcf46f-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/46a079c3-8654-4ed5-9466-088320bda559-m0.avro, hdfs:/tmp/tables/job_metrics_tmp/metadata/f239498c-7386-4f31-8421-518105ffbf6a-m0.avro]}" ] }, "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], "source": [ "val tables = new HadoopTables(spark.sessionState.newHadoopConf())\n", "val table = tables.load(path)\n", "\n", "table.currentSnapshot" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "7087" ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "import scala.collection.JavaConverters._\n", "\n", "table.currentSnapshot.addedFiles.asScala.size" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [], "source": [ "table.newAppend.commit // use a merge commit to create a single manifest" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "BaseSnapshot{id=1515605215920, timestamp_ms=1515605220253, manifests=[hdfs:/tmp/tables/job_metrics_tmp/metadata/213364b0-d97f-49bf-9126-7273b9784cfb-m0.avro]}" ] }, "execution_count": 9, "metadata": {}, "output_type": "execute_result" } ], "source": [ "table.currentSnapshot" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "hide_input": false, "kernelspec": { "display_name": "Spark 2.0.0 - Scala 2.11", "language": "scala", "name": "spark2-scala" }, "language_info": { "codemirror_mode": "text/x-scala", "file_extension": ".scala", "mimetype": "text/x-scala", "name": "scala", "pygments_lexer": "scala", "version": "2.11.8" } }, "nbformat": 4, "nbformat_minor": 2 }