Camus Compressor

Camus Compressor merges files created by Camus and saves them in a compressed format.

Camus is massively used at Allegro for dumping more than 200 Kafka topics onto HDFS. The script runs every 15 minutes and creates one file per Kafka partition which results in about 76800 small files per day. Most of the files do not exceed Hadoop block size. This is a clear Hadoop antipattern which leads to performance issues, for example extensive number of mappers in SQL queries’ executions.

Camus Compressor solves this issue by merging files within Hive partition and compressing them. It does not change Camus directories structure and supports well daily and hourly partitioning. The tool runs in YARN and is build on Spark.

Supported compressors

• Snappy
• LZO (needs extra hadoop-lzo package on nodes)
• Deflate
• None (without compression, files are compacted only).

How to use

As mentioned above You need Spark packages to run Camus Compressor. Provided src/main/resources/compressor.sh file helps executing spark-submit commands by setting options:

• P: Configuration file path (default: /etc/camus-compressor/camus-compressor.properties);
• q: YARN queue name (default: default);
• e: Parrallelism level of task (number of spark executors, default: 2);
• m: Spark --master option;
• c: Spark --conf option. You can use it to pass extra arguments to job.
• d: Spark --driver-memory option. If you have huge number of partitions to compress, please consider set this option to, for example, 4g.

Configuration file

In a configuration file (/etc/camus-compressor/camus-compressor.properties) you can set following options:

• spark.compressor.input.format - input files format (avro or json)
• spark.compressor.processing.mode - processing mode, can be either:
  • all: compress all camus dir, put main camus dir as input path
  • topic: compress only one topic, put topic dir as the input path
  • unit: compress low-level camus directory (hour dir on hourly patitioning and day dir on daily partitioning), put appropriate input path to unit
• spark.compressor.input.path - path to the uncompressed file
  • in all mode is a path to the directory containing topics that should be compressed (/user/username/topics)
  • in topic mode is a path to the topic directory (/user/username/topics/my_topic)
  • in unit mode is a path to one partition (directory containing files: /user/username/topics/my_topic/daily/2016/03/23/)
• spark.compressor.output.compression - output compression (snappy, deflate, none, lzo).
• spark.compressor.avro.schema.repository.class - currently only schema-repo is supported. Use pl.allegro.tech.hadoop.compressor.schema.SchemaRepoSchemaRepository as a value.
• spark.compressor.avro.schema.repository.url - URL to the schema-repo, for example: http://example.com/schema-repo.
• spark.compressor.processing.topic-name-retriever.class:
  • use pl.allegro.tech.hadoop.compressor.schema.KafkaTopicNameRetriever if directories on HDFS are named after Kafka topics with dots replaced with underscores (topic.name on Kafka would be topic_name on HDFS).
  • use pl.allegro.tech.hadoop.compressor.schema.IdentityTopicNameRetriever if directories on HDFS have names identical to topic names.
• spark.compressor.zookeeper.paths - when using KafkaTopicNameRetriever, please provide Zookeeper connection string.
• spark.compressor.processing.delay - compression delay, in days (compress data older than given number of days, default: 2).
• spark.compressor.processing.mode.all.excludes - comma separated list of directories excluded from compression. If you want to blacklist some topics, you can pass their names here.
• spark.compressor.processing.mode.all.timeout.minutes - after this period of time Compressor will be termineted.
• spark.compressor.processing.mode.topic.pattern - comma separated list of directories that contains data that need to be compressed. For Camus you can use hourly,daily.
• spark.compressor.processing.force - by default Camus Compressor doesn't compress directories again. Set this option to true to force recompression.
• spark.compressor.processing.working.dir - directory where temprary files will be stored.
• spark.compressor.processing.calculate.counts - if set to true, Compressor will calculate counts for every unit (i.e. partition) before and after compression. If they are not equal, it will throws an exception and leaves compressed files in working.dir.
• spark.executors.instances - number of topics concurrently processed.

How to build

Camus Compressor is shipped as fatjar file or Debian package and build using Gradle. To build fat-jar run in build/libs/:

./gradlew shadowJar

To build debian package camus-compressor in build/, run:

./gradlew shadowJar prepareControlFile buildDeb

Sample usage

Before executing compressor.sh script, make sure that SPARK_SUBMIT variable is set to the spark-submit location of Spark in version 1.6.1 or above.

export SPARK_SUBMIT="/usr/bin/spark-submit"
compressor.sh -P /etc/camus-compressor/camus-compressor.properties \
    -e 10 \
    -q default \
    -d 4g \
    -m yarn-cluster \
    -c "spark.executor.extraJavaOptions=-XX:+PrintGCDetails -XX:+PrintGCTimeStamps"

License

Copyright 2015 Allegro Group

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.