Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance. Project price only 1 $
You can buy this project and download/modify it how often you want.
/*
*
* * Licensed to the Apache Software Foundation (ASF) under one or more
* * contributor license agreements. See the NOTICE file distributed with
* * this work for additional information regarding copyright ownership.
* * The ASF licenses this file to You 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.
*
*
*/
package com.ebiznext.comet.config
import java.io.ObjectStreamException
import java.lang.management.{ManagementFactory, RuntimeMXBean}
import java.util.concurrent.TimeUnit
import java.util.{Locale, UUID, Map => juMap}
import com.ebiznext.comet.schema.handlers.{
AirflowLauncher,
HdfsStorageHandler,
LaunchHandler,
SimpleLauncher
}
import com.ebiznext.comet.schema.model.Sink
import com.ebiznext.comet.utils.{CometObjectMapper, Version}
import com.fasterxml.jackson.annotation.JsonIgnore
import com.fasterxml.jackson.databind.ObjectMapper
import com.typesafe.config.{Config, ConfigValueFactory}
import com.typesafe.scalalogging.{Logger, StrictLogging}
import configs.Configs
import configs.syntax._
import org.apache.spark.storage.StorageLevel
import org.slf4j.MDC
import scala.concurrent.duration.FiniteDuration
object Settings extends StrictLogging {
private def loggerForCompanionInstances: Logger = logger
/** @param endpoint : Airflow REST API endpoint, aka. http://127.0.0.1:8080/api/experimental
*/
final case class Airflow(endpoint: String, ingest: String)
/** datasets in the data pipeline go through several stages and
* are stored on disk at each of these stages.
* This setting allow to customize the folder names of each of these stages.
*
* @param pending : Name of the pending area
* @param unresolved : Named of the unresolved area
* @param archive : Name of the archive area
* @param ingesting : Name of the ingesting area
* @param accepted : Name of the accepted area
* @param rejected : Name of the rejected area
* @param business : Name of the business area
*/
final case class Area(
pending: String,
unresolved: String,
archive: String,
ingesting: String,
accepted: String,
rejected: String,
business: String
) {
val acceptedFinal: String = accepted.toLowerCase(Locale.ROOT)
val rejectedFinal: String = rejected.toLowerCase(Locale.ROOT)
val businessFinal: String = business.toLowerCase(Locale.ROOT)
}
/** @param options : Map of privacy algorightms name -> PrivacyEngine
*/
final case class Privacy(options: juMap[String, String])
final case class Elasticsearch(active: Boolean, options: juMap[String, String])
/** @param discreteMaxCardinality : Max number of unique values allowed in cardinality compute
*/
final case class Metrics(
path: String,
discreteMaxCardinality: Int,
active: Boolean,
sink: Sink
)
final case class Audit(
path: String,
sink: Sink,
maxErrors: Int
)
/** Describes a connection to a JDBC-accessible database engine
*
* @param format source / sink format (jdbc by default). Cf spark.format possible values
* @param mode Spark SaveMode to use. If not present, the save mode will be computed from the write disposition set in the YAM file
* @param options any option required by the format used to ingest / tranform / compute the data. Eg for JDBC uri, user and password are required
* uri the URI of the database engine. It must start with "jdbc:"
* user the username under which to connect to the database engine
* password the password to use in order to connect to the database engine
* @param engineOverride the index into the [[Comet.jdbcEngines]] map of the underlying database engine, in case
* one cannot use the engine name from the uri
*
* @note the use case for engineOverride is when you need to have an alternate schema definition
* (e.g. non-standard table names) alongside with the regular schema definition, on the same
* underlying engine.
*/
final case class Connection(
format: String = "jdbc",
mode: Option[String] = None,
options: Map[String, String] = Map.empty,
engineOverride: Option[String] = None
) {
def engine: String = engineOverride.getOrElse(options("url").split(':')(1))
}
/** Describes how to use a specific type of JDBC-accessible database engine
*
* @param tables for each of the Standard Table Names used by Comet, the specific SQL DDL statements as expected
* in the engine's own dialect.
*/
final case class JdbcEngine(
tables: scala.collection.Map[String, JdbcEngine.TableDdl]
)
object JdbcEngine {
/** A descriptor of the specific SQL DDL statements required to manage a specific Comet table in a JDBC-accessible
* database engine
*
* @param createSql the SQL Create Table statement with the database-specific type, constraints etc. tacked on.
* @param pingSql a cheap SQL query whose results are irrelevant but guaranteed to trigger an error in case the table is absent
*
* @note pingSql is optional, and will default to `select * from `name` where 1=0` as Spark SQL does
*/
final case class TableDdl(createSql: String, pingSql: Option[String] = None) {
def effectivePingSql(tableName: String): String =
pingSql.getOrElse(s"select * from $tableName where 1=0")
}
}
final case class Lock(
path: String,
metricsTimeout: Long,
ingestionTimeout: Long,
pollTime: FiniteDuration = FiniteDuration(5000L, TimeUnit.MILLISECONDS),
refreshTime: FiniteDuration = FiniteDuration(5000L, TimeUnit.MILLISECONDS)
)
final case class Atlas(uri: String, user: String, password: String, owner: String)
final case class Internal(cacheStorageLevel: StorageLevel) {}
/** @param datasets : Absolute path, datasets root folder beneath which each area is defined.
* @param metadata : Absolute path, location where all types / domains and auto jobs are defined
* @param metrics : Absolute path, location where all computed metrics are stored
* @param audit : Absolute path, location where all log are stored
* @param archive : Should we backup the ingested datasets ? true by default
* @param defaultWriteFormat : Choose between parquet, orc ... Default is parquet
* @param launcher : Cron Job Manager : simple (useful for testing) or airflow ? simple by default
* @param analyze : Should we create basics Hive statistics on the generated dataset ? true by default
* @param hive : Should we create a Hive Table ? true by default
* @param area : see Area above
* @param airflow : Airflow end point. Should be defined even if simple launccher is used instead of airflow.
*/
final case class Comet(
tmpdir: String,
jobId: String,
datasets: String,
metadata: String,
metrics: Metrics,
audit: Audit,
archive: Boolean,
lock: Lock,
defaultWriteFormat: String,
csvOutput: Boolean,
privacyOnly: Boolean,
launcher: String,
chewerPrefix: String,
rowValidatorClass: String,
analyze: Boolean,
hive: Boolean,
grouped: Boolean,
mergeForceDistinct: Boolean,
area: Area,
airflow: Airflow,
elasticsearch: Elasticsearch,
hadoop: juMap[String, String],
connections: Map[String, Connection],
jdbcEngines: Map[String, JdbcEngine],
atlas: Atlas,
privacy: Privacy,
fileSystem: Option[String],
metadataFileSystem: Option[String],
internal: Option[Internal],
udfs: Option[String]
) extends Serializable {
@JsonIgnore
def isElasticsearchSupported(): Boolean = {
if (
Version(util.Properties.versionNumberString).compareTo(Version("2.12")) >= 0
&& elasticsearch.active
) {
logger.warn("""Elasticsearch inserts won't be effective before es-hadoop support scala 2.12
|See https://github.com/elastic/elasticsearch-hadoop/pull/1308
|""".stripMargin)
false
} else true
}
val cacheStorageLevel =
internal.map(_.cacheStorageLevel).getOrElse(StorageLevel.MEMORY_AND_DISK)
@throws(classOf[ObjectStreamException])
protected def writeReplace: AnyRef = {
Comet.JsonWrapped(this)
}
}
object Comet {
private case class JsonWrapped(jsonValue: String) {
@throws(classOf[ObjectStreamException])
protected def readResolve: AnyRef = {
val unwrapped = JsonWrapped.jsonMapper.readValue(jsonValue, classOf[Comet])
unwrapped
}
}
private object JsonWrapped {
private def jsonMapper: ObjectMapper = new CometObjectMapper()
def apply(comet: Comet): JsonWrapped = {
val writer = jsonMapper.writerFor(classOf[Comet])
val asJson = writer.writeValueAsString(comet)
JsonWrapped(asJson)
}
}
}
private implicit val sinkConfigs: Configs[Sink] = Configs.derive[Sink]
private implicit val jdbcEngineConfigs: Configs[JdbcEngine] = Configs.derive[JdbcEngine]
private implicit val storageLevelConfigs: Configs[StorageLevel] =
Configs[String].map(StorageLevel.fromString)
def apply(config: Config): Settings = {
val jobId = UUID.randomUUID().toString
val effectiveConfig = config
.withValue("job-id", ConfigValueFactory.fromAnyRef(jobId, "per JVM instance"))
val loaded = effectiveConfig.extract[Comet].valueOrThrow { error =>
error.messages.foreach(err => logger.error(err))
throw new Exception(s"Failed to load config: $error")
}
logger.info(s"Using Config $loaded")
Settings(loaded, effectiveConfig.getConfig("spark"))
}
val cometInputFileNameColumn: String = "comet_input_file_name"
}
/** This class holds the current Comet settings and an assembly of reference instances for core, shared services
*
* SMELL: this may be the start of a Dependency Injection root (but at 2-3 objects, is DI justified? probably not
* quite yet) — cchepelov
*/
final case class Settings(comet: Settings.Comet, sparkConfig: Config) {
def logger: Logger = Settings.loggerForCompanionInstances
@transient
lazy val storageHandler: HdfsStorageHandler = {
implicit val self: Settings =
this /* TODO: remove this once HdfsStorageHandler explicitly takes Settings or Settings.Comet in */
new HdfsStorageHandler(comet.fileSystem)
}
@transient
lazy val metadataStorageHandler: HdfsStorageHandler = {
implicit val self: Settings =
this /* TODO: remove this once HdfsStorageHandler explicitly takes Settings or Settings.Comet in */
new HdfsStorageHandler(comet.metadataFileSystem)
}
@transient
lazy val launcherService: LaunchHandler = comet.launcher match {
case "simple" => new SimpleLauncher()
case "airflow" => new AirflowLauncher()
}
/** Publish MDC information into the logging stack.
*
* @note this is inherently an effectful operation, which effects global shared mutable state.
* It should make little sense to run this code in tests.
*/
def publishMDCData(): Unit = {
val rt: RuntimeMXBean = ManagementFactory.getRuntimeMXBean
MDC.put("PID", rt.getName)
val oldJobId = Option(MDC.get("JID")).getOrElse(comet.jobId)
require(
oldJobId == comet.jobId,
s"cannot publish different MDC data; a previous jobId ${oldJobId} had been published," +
s" attempting to reset to ${comet.jobId}"
)
MDC.put("JID", comet.jobId)
}
}
