Embulk, an open-source plugin-based parallel bulk data loader
•
143 likes•238,441 views
The document discusses Embulk, an open-source parallel bulk data loader that uses plugins. Embulk loads records from various sources ("A") to various targets ("B") using plugins for different source and target types. This makes the painful process of data integration more relaxed. Embulk executes in parallel, validates data, handles errors, behaves deterministically, and allows for idempotent retries of bulk loads.
Report
Share
Recommended for you
Recommended for you
Recommended for you
Recommended for you
Recommended for you
Recommended for you
Recommended for you
Recommended for you
More Related Content
What's hot
What's hot (20)
Viewers also liked
Viewers also liked (19)
Similar to Embulk, an open-source plugin-based parallel bulk data loader
Similar to Embulk, an open-source plugin-based parallel bulk data loader (20)
More from Sadayuki Furuhashi
More from Sadayuki Furuhashi (20)
Recently uploaded
Recently uploaded (20)
Embulk, an open-source plugin-based parallel bulk data loader
- 1. Sadayuki Furuhashi Founder & Software Architect Treasure Data, inc. EmbulkAn open-source plugin-based parallel bulk data loader that makes painful data integration work relaxed. Sharing our knowledge on RubyGems to manage arbitrary files.
- 2. A little about me... > Sadayuki Furuhashi > github/twitter: @frsyuki > Treasure Data, Inc. > Founder & Software Architect > Open-source hacker > MessagePack - Efficient object serializer > Fluentd - An unified data collection tool > Prestogres - PostgreSQL protocol gateway for Presto > Embulk - A plugin-based parallel bulk data loader > ServerEngine - A Ruby framework to build multiprocess servers > LS4 - A distributed object storage with cross-region replication > kumofs - A distributed strong-consistent key-value data store
- 3. Today’s talk > What’s Embulk? > How Embulk works? > The architecture > Writing Embulk plugins > Roadmap & Development > Q&A + Discussion
- 4. What’s Embulk? > An open-source parallel bulk data loader > using plugins > to make data integration relaxed.
- 5. What’s Embulk? > An open-source parallel bulk data loader > loads records from “A” to “B” > using plugins > for various kinds of “A” and “B” > to make data integration relaxed. > which was very painful… Storage, RDBMS, NoSQL, Cloud Service, etc. broken records, transactions (idempotency), performance, …
- 6. The pains of bulk data loading Example: load a 10GB CSV file to PostgreSQL > 1. First attempt → fails > 2. Write a script to make the records cleaned • Convert ”20150127T190500Z” → “2015-01-27 19:05:00 UTC” • Convert “N" → “” • many cleanings… > 3. Second attempt → another error • Convert “Inf” → “Infinity” > 4. Fix the script, retry, retry, retry… > 5. Oh, some data got loaded twice!?
- 7. The pains of bulk data loading Example: load a 10GB CSV file to PostgreSQL > 6. Ok, the script worked. > 7. Register it to cron to sync data every day. > 8. One day… it fails with another error • Convert invalid UTF-8 byte sequence to U+FFFD
- 8. The pains of bulk data loading Example: load 10GB CSV × 720 files > Most of scripts are slow. • People have little time to optimize bulk load scripts > One file takes 1 hour → 720 files takes 1 month (!?) A lot of integration efforts for each storages: > XML, JSON, Apache log format (+some custom), … > SAM, BED, BAI2, HDF5, TDE, SequenceFile, RCFile… > MongoDB, Elasticsearch, Redshift, Salesforce, …
- 9. The problems: > Data cleaning (normalization) > How to normalize broken records? > Error handling > How to remove broken records? > Idempotent retrying > How to retry without duplicated loading? > Performance optimization > How to optimize the code or parallelize?
- 10. The problems at Treasure Data Treasure Data Service? > “Fast, powerful SQL access to big data from connected applications and products, with no new infrastructure or special skills required.” > Customers want to try Treasure Data, but > SEs write scripts to bulk load their data. Hard work :( > Customers want to migrate their big data, but > Hard work :( > Fluentd solved streaming data collection, but > bulk data loading is another problem.
- 11. A solution: > Package the efforts as a plugin. > data cleaning, error handling, retrying > Share & reuse the plugin. > don’t repeat the pains! > Keep improving the plugin code. > rather than throwing away the efforts every time > using OSS-style pull-reqs & frequent releases.
- 12. Embulk Embulk is an open-source, plugin-based parallel bulk data loader that makes data integration works relaxed.
- 13. HDFS MySQL Amazon S3 Embulk CSV Files SequenceFile Salesforce.com Elasticsearch Cassandra Hive Redis
- 14. HDFS MySQL Amazon S3 Embulk CSV Files SequenceFile Salesforce.com Elasticsearch Cassandra Hive Redis ✓ Parallel execution ✓ Data validation ✓ Error recovery ✓ Deterministic behavior ✓ Idempotet retrying bulk load
- 15. HDFS MySQL Amazon S3 Embulk CSV Files SequenceFile Salesforce.com Elasticsearch Cassandra Hive Redis ✓ Parallel execution ✓ Data validation ✓ Error recovery ✓ Deterministic behavior ✓ Idempotet retrying Plugins Plugins bulk load
- 17. # install $ wget https://bintray.com/artifact/download/ embulk/maven/embulk-0.2.0.jar -o embulk.jar $ chmod 755 embulk.jar Installing embulk Bintray releases Embulk is released on Bintray wget embulk.jar
- 18. # install $ wget https://bintray.com/artifact/download/ embulk/maven/embulk-0.2.0.jar -o embulk.jar $ chmod 755 embulk.jar # guess $ vi partial-config.yml $ ./embulk guess partial-config.yml -o config.yml Guess format & schema in: type: file paths: [data/examples/] out: type: example in: type: file paths: [data/examples/] decoders: - {type: gzip} parser: charset: UTF-8 newline: CRLF type: csv delimiter: ',' quote: '"' header_line: true columns: - name: time type: timestamp format: '%Y-%m-%d %H:%M:%S' - name: account type: long - name: purchase type: timestamp format: '%Y%m%d' - name: comment type: string out: type: example guess by guess plugins
- 19. # install $ wget https://bintray.com/artifact/download/ embulk/maven/embulk-0.2.0.jar -o embulk.jar $ chmod 755 embulk.jar # guess $ vi partial-config.yml $ ./embulk guess partial-config.yml -o config.yml # preview $ ./embulk preview config.yml $ vi config.yml # if necessary +--------------------------------------+---------------+--------------------+ | time:timestamp | uid:long | word:string | +--------------------------------------+---------------+--------------------+ | 2015-01-27 19:23:49 UTC | 32,864 | embulk | | 2015-01-27 19:01:23 UTC | 14,824 | jruby | | 2015-01-28 02:20:02 UTC | 27,559 | plugin | | 2015-01-29 11:54:36 UTC | 11,270 | fluentd | +--------------------------------------+---------------+--------------------+ Preview & fix config
- 20. # install $ wget https://bintray.com/artifact/download/ embulk/maven/embulk-0.2.0.jar -o embulk.jar $ chmod 755 embulk.jar # guess $ vi partial-config.yml $ ./embulk guess partial-config.yml -o config.yml # preview $ ./embulk preview config.yml $ vi config.yml # if necessary # run $ ./embulk run config.yml -o config.yml in: type: file paths: [data/examples/] decoders: - {type: gzip} parser: charset: UTF-8 newline: CRLF type: csv delimiter: ',' quote: '"' header_line: true columns: - name: time type: timestamp format: '%Y-%m-%d %H:%M:%S' - name: account type: long - name: purchase type: timestamp format: '%Y%m%d' - name: comment type: string last_paths: [data/examples/sample_001.csv.gz] out: type: example Deterministic run
- 21. in: type: file paths: [data/examples/] decoders: - {type: gzip} parser: charset: UTF-8 newline: CRLF type: csv delimiter: ',' quote: '"' header_line: true columns: - name: time type: timestamp format: '%Y-%m-%d %H:%M:%S' - name: account type: long - name: purchase type: timestamp format: '%Y%m%d' - name: comment type: string last_paths: [data/examples/sample_002.csv.gz] out: type: example Repeat # install $ wget https://bintray.com/artifact/download/ embulk/maven/embulk-0.2.0.jar -o embulk.jar $ chmod 755 embulk.jar # guess $ vi partial-config.yml $ ./embulk guess partial-config.yml -o config.yml # preview $ ./embulk preview config.yml $ vi config.yml # if necessary # run $ ./embulk run config.yml -o config.yml # repeat $ ./embulk run config.yml -o config.yml $ ./embulk run config.yml -o config.yml
- 22. The architecture
- 23. InputPlugin OutputPlugin Embulk executor plugin read records write records
- 24. InputPlugin OutputPlugin Embulk executor plugin MySQL, Cassandra, HBase, Elasticsearch, Treasure Data, … record record
- 25. InputPlugin FileInputPlugin OutputPlugin FileOutputPlugin EncoderPlugin FormatterPlugin DecoderPlugin ParserPlugin Embulk executor plugin read files decompress parse files into records write files compress format records into files
- 26. InputPlugin FileInputPlugin OutputPlugin FileOutputPlugin EncoderPlugin FormatterPlugin DecoderPlugin ParserPlugin Embulk executor plugin HDFS, S3, Riak CS, … gzip, bzip2, 3des, … CSV, JSON, RCFile, … buffer buffer record record buffer buffer
- 28. InputPlugin module Embulk class InputExample < InputPlugin Plugin.register_input('example', self) def self.transaction(config, &control) # read config task = { 'message' => config.param('message', :string, default: nil) } threads = config.param('threads', :int, default: 2) columns = [ Column.new(0, 'col0', :long), Column.new(1, 'col1', :double), Column.new(2, 'col2', :string), ] # BEGIN here commit_reports = yield(task, columns, threads) # COMMIT here puts "Example input finished" return {} end def run(task, schema, index, page_builder) puts "Example input thread #{@index}…" 10.times do |i| @page_builder.add([i, 10.0, "example"]) end @page_builder.finish commit_report = { } return commit_report end end end
- 29. OutputPlugin module Embulk class OutputExample < OutputPlugin Plugin.register_output('example', self) def self.transaction( config, schema, processor_count, &control) # read config task = { 'message' => config.param('message', :string, default: "record") } puts "Example output started." commit_reports = yield(task) puts "Example output finished. Commit reports = #{commit_reports.to_json}" return {} end def initialize(task, schema, index) puts "Example output thread #{index}..." super @message = task.prop('message', :string) @records = 0 end def add(page) page.each do |record| hash = Hash[schema.names.zip(record)] puts "#{@message}: #{hash.to_json}" @records += 1 end end def finish end def abort end def commit commit_report = { "records" => @records } return commit_report end end end
- 30. GuessPlugin # guess_gzip.rb module Embulk class GzipGuess < GuessPlugin Plugin.register_guess('gzip', self) GZIP_HEADER = "x1f x8b".force_encoding('ASCII-8BIT').freeze def guess(config, sample_buffer) if sample_buffer[0,2] == GZIP_HEADER return {"decoders" => [{"type" => "gzip"}]} end return {} end end end # guess_ module Embulk class GuessNewline < TextGuessPlugin Plugin.register_guess('newline', self) def guess_text(config, sample_text) cr_count = sample_text.count("r") lf_count = sample_text.count("n") crlf_count = sample_text.scan(/rn/).length if crlf_count > cr_count / 2 && crlf_count > lf_count / 2 return {"parser" => {"newline" => "CRLF"}} elsif cr_count > lf_count / 2 return {"parser" => {"newline" => "CR"}} else return {"parser" => {"newline" => "LF"}} end end end end
- 31. Releasing to RubyGems Examples > embulk-plugin-postgres-json.gem > https://github.com/frsyuki/embulk-plugin-postgres-json > embulk-plugin-redis.gem > https://github.com/komamitsu/embulk-plugin-redis > embulk-plugin-input-sfdc-event-log-files.gem > https://github.com/nahi/embulk-plugin-input-sfdc-event- log-files
- 33. Roadmap > Add missing JRuby Plugin APIs > ParserPlugin, FormatterPlugin > DecoderPlugin, EncoderPlugin > Add Executor plugin SPI > Add ssh distributed executor > embulk run —command ssh %host embulk run %task > Add MapReduce executor > Add support for nested records (?)
- 34. Contributing to the Embulk project > Pull-requests & issues on Github > Posting blogs > “I tried Embulk. Here is how it worked” > “I read Embulk code. Here is how it’s written” > “Embulk is good because…but bad because…” > Talking on Twitter with a word “embulk" > Writing & releasing plugins > Windows support > Integration to other software > ETL tools, Fluentd, Hadoop, Presto, …
- 35. Q&A + Discussion? Hiroshi Nakamura @nahi Muga Nishizawa @muga_nishizawa Sadayuki Furuhashi @frsyuki Embulk committers:
- 36. https://jobs.lever.co/treasure-data Cloud service for the entire data pipeline. We’re hiring!