HDInsight에서 Azure PowerShell을 사용하여 Apache Sqoop 작업 실행

Azure PowerShell을 사용하여 Azure HDInsight에서 Apache Sqoop 작업을 실행하여 HDInsight 클러스터와 Azure SQL Database 또는 SQL Server 간에 데이터를 가져오고 내보내는 방법에 대해 알아봅니다. 이 문서는 HDInsight에서 Hadoop과 함께 Apache Sqoop 사용의 연속입니다.

필수 조건

Sqoop 내보내기

Hive에서 SQL로 내보냅니다.

이 예제에서는 Hive hivesampletable 테이블의 데이터를 SQL의 mobiledata 테이블로 내보냅니다. 아래 변수 값을 설정한 다음, 명령을 실행합니다.

$hdinsightClusterName = ""
$httpPassword = ''
$sqlDatabasePassword = ''

# These values only need to be changed if the template was not followed.
$httpUserName = "admin"
$sqlServerLogin = "sqluser"
$sqlServerName = $hdinsightClusterName + "dbserver"
$sqlDatabaseName = $hdinsightClusterName + "db"

$pw = ConvertTo-SecureString -String $httpPassword -AsPlainText -Force
$httpCredential = New-Object System.Management.Automation.PSCredential($httpUserName,$pw)

# Connection string
$connectionString = "jdbc:sqlserver://$sqlServerName.database.windows.net;user=$sqlServerLogin@$sqlServerName;password=$sqlDatabasePassword;database=$sqlDatabaseName"

# start export
New-AzHDInsightSqoopJobDefinition `
    -Command "export --connect $connectionString --table mobiledata --hcatalog-table hivesampletable" `
    | Start-AzHDInsightJob `
        -ClusterName $hdinsightClusterName `
        -HttpCredential $httpCredential

대체 실행

  1. 아래 코드는 동일한 내보내기를 수행합니다. 그러나 출력 로그를 읽는 방법을 제공합니다. 코드를 실행하여 내보내기를 시작합니다.

    $sqoopCommand = "export --connect $connectionString --table mobiledata --hcatalog-table hivesampletable"
    
    $sqoopDef = New-AzHDInsightSqoopJobDefinition `
        -Command $sqoopCommand
    
    $sqoopJob = Start-AzHDInsightJob `
                    -ClusterName $hdinsightClusterName `
                    -HttpCredential $httpCredential `
                    -JobDefinition $sqoopDef
    
  2. 아래 코드는 출력 로그를 표시합니다. 아래 코드를 실행합니다.

    Get-AzHDInsightJobOutput `
        -ClusterName $hdinsightClusterName `
        -HttpCredential $httpCredential `
        -JobId $sqoopJob.JobId `
        -DisplayOutputType StandardError
    
    Get-AzHDInsightJobOutput `
        -ClusterName $hdinsightClusterName `
        -HttpCredential $httpCredential `
        -JobId $sqoopJob.JobId `
        -DisplayOutputType StandardOutput
    

The specified blob does not exist. 오류 메시지가 표시되면 몇 분 후에 다시 시도하세요.

Sqoop 가져오기

SQL에서 Azure Storage로 이 예제에서는 SQL의 mobiledata 테이블에서 HDInsight의 wasb:///tutorials/usesqoop/importeddata 디렉터리로 데이터를 가져옵니다. 데이터의 필드는 탭 문자로 구분되어 있으며 줄은 줄 바꿈 문자로 종료됩니다. 이 예제에서는 이전 예제를 완료했다고 가정합니다.

$sqoopCommand = "import --connect $connectionString --table mobiledata --target-dir wasb:///tutorials/usesqoop/importeddata --fields-terminated-by '\t' --lines-terminated-by '\n' -m 1"


$sqoopDef = New-AzHDInsightSqoopJobDefinition `
    -Command $sqoopCommand

$sqoopJob = Start-AzHDInsightJob `
                -ClusterName $hdinsightClusterName `
                -HttpCredential $httpCredential `
                -JobDefinition $sqoopDef

Get-AzHDInsightJobOutput `
    -ClusterName $hdinsightClusterName `
    -HttpCredential $httpCredential `
    -JobId $sqoopJob.JobId `
    -DisplayOutputType StandardError

Get-AzHDInsightJobOutput `
    -ClusterName $hdinsightClusterName `
    -HttpCredential $httpCredential `
    -JobId $sqoopJob.JobId `
    -DisplayOutputType StandardOutput

추가 Sqoop 내보내기 예

이는 기본 스토리지 계정에서 /tutorials/usesqoop/data/sample.log 데이터를 내보낸 다음, SQL Server 데이터베이스의 log4jlogs라는 테이블로 가져오는 강력한 예제입니다. 이 예제는 이전 예제에 영향을 받지 않습니다.

다음 PowerShell 스크립트는 원본 파일을 전처리한 후 테이블 log4jlogs로 내보냅니다. CLUSTERNAME, CLUSTERPASSWORDSQLPASSWORD를 필수 구성 요소에서 사용한 값으로 바꿉니다.

<#------ BEGIN USER INPUT ------#>
$hdinsightClusterName = "CLUSTERNAME"
$httpUserName = "admin"  #default is admin, update as needed
$httpPassword = 'CLUSTERPASSWORD'
$sqlDatabasePassword = 'SQLPASSWORD'
<#------- END USER INPUT -------#>

# Other fixed variable that should be used as is
$sqlServerName = $hdinsightClusterName + "dbserver"
$sqlDatabaseName = $hdinsightClusterName + "db"
$tableName_log4j = "log4jlogs"
$exportDir_log4j = "/tutorials/usesqoop/data"
$sourceBlobName = "example/data/sample.log"
$destBlobName = "tutorials/usesqoop/data/sample.log"
$sqljdbcdriver = "/user/oozie/share/lib/sqoop/mssql-jdbc-7.0.0.jre8.jar"

$cluster = Get-AzHDInsightCluster -ClusterName $hdinsightClusterName
$defaultStorageAccountName = $cluster.DefaultStorageAccount -replace '.blob.core.windows.net'
$defaultStorageContainer = $cluster.DefaultStorageContainer
$resourceGroup = $cluster.ResourceGroup

$sqlServer = Get-AzSqlServer -ResourceGroupName $resourceGroup -ServerName $sqlServerName
$sqlServerLogin = $sqlServer.SqlAdministratorLogin
$sqlServerFQDN = $sqlServer.FullyQualifiedDomainName

#Connect to Azure subscription
Write-Host "`nConnecting to your Azure subscription ..." -ForegroundColor Green
try{Get-AzContext}
catch{Connect-AzAccount}

#pre-process the source file
Write-Host "`nPreprocessing the source file ..." -ForegroundColor Green

# This procedure creates a new file with $destBlobName
# Define the connection string
$defaultStorageAccountKey = (Get-AzStorageAccountKey `
                                -ResourceGroupName $resourceGroup `
                                -Name $defaultStorageAccountName)[0].Value

# Create block blob objects referencing the source and destination blob.
$storageAccount = Get-AzStorageAccount `
    -ResourceGroupName $resourceGroup `
    -Name $defaultStorageAccountName

$storageContainer = ($storageAccount |Get-AzStorageContainer -Name $defaultStorageContainer).CloudBlobContainer

$sourceBlob = $storageContainer.GetBlockBlobReference($sourceBlobName)
$destBlob = $storageContainer.GetBlockBlobReference($destBlobName)

# Define a MemoryStream and a StreamReader for reading from the source file
$stream = New-Object System.IO.MemoryStream
$stream = $sourceBlob.OpenRead()
$sReader = New-Object System.IO.StreamReader($stream)

# Define a MemoryStream and a StreamWriter for writing into the destination file
$memStream = New-Object System.IO.MemoryStream
$writeStream = New-Object System.IO.StreamWriter $memStream

# Pre-process the source blob
$exString = "java.lang.Exception:"
while(-Not $sReader.EndOfStream){
    $line = $sReader.ReadLine()
    $split = $line.Split(" ")

    # remove the "java.lang.Exception" from the first element of the array
    # for example: java.lang.Exception: 2012-02-03 19:11:02 SampleClass8 [WARN] problem finding id 153454612
    if ($split[0] -eq $exString){
        #create a new ArrayList to remove $split[0]
        $newArray = [System.Collections.ArrayList] $split
        $newArray.Remove($exString)

        # update $split and $line
        $split = $newArray
        $line = $newArray -join(" ")
    }

    # remove the lines that has less than 7 elements
    if ($split.count -ge 7){
        write-host $line
        $writeStream.WriteLine($line)
    }
}

# Write to the destination blob
$writeStream.Flush()
$memStream.Seek(0, "Begin")
$destBlob.UploadFromStream($memStream)

#export the log file from the cluster to SQL
Write-Host "Exporting the log file ..." -ForegroundColor Green

$pw = ConvertTo-SecureString -String $httpPassword -AsPlainText -Force
$httpCredential = New-Object System.Management.Automation.PSCredential($httpUserName,$pw)

# Connection string
$connectionString = "jdbc:sqlserver://$sqlServerFQDN;user=$sqlServerLogin@$sqlServerName;password=$sqlDatabasePassword;database=$sqlDatabaseName"

# Submit a Sqoop job
$sqoopDef = New-AzHDInsightSqoopJobDefinition `
    -Command "export --connect $connectionString --table $tableName_log4j --export-dir $exportDir_log4j --input-fields-terminated-by \0x20 -m 1" `
    -Files $sqljdbcdriver

$sqoopJob = Start-AzHDInsightJob `
                -ClusterName $hdinsightClusterName `
                -HttpCredential $httpCredential `
                -JobDefinition $sqoopDef

Wait-AzHDInsightJob `
    -ResourceGroupName $resourceGroup `
    -ClusterName $hdinsightClusterName `
    -HttpCredential $httpCredential `
    -JobId $sqoopJob.JobId

Write-Host "Standard Error" -BackgroundColor Green
Get-AzHDInsightJobOutput `
    -ResourceGroupName $resourceGroup `
    -ClusterName $hdinsightClusterName `
    -DefaultStorageAccountName $defaultStorageAccountName `
    -DefaultStorageAccountKey $defaultStorageAccountKey `
    -DefaultContainer $defaultStorageContainer `
    -HttpCredential $httpCredential `
    -JobId $sqoopJob.JobId `
    -DisplayOutputType StandardError

Write-Host "Standard Output" -BackgroundColor Green
Get-AzHDInsightJobOutput `
    -ResourceGroupName $resourceGroupName `
    -ClusterName $hdinsightClusterName `
    -DefaultStorageAccountName $defaultStorageAccountName `
    -DefaultStorageAccountKey $defaultStorageAccountKey `
    -DefaultContainer $defaultStorageContainer `
    -HttpCredential $httpCredential `
    -JobId $sqoopJob.JobId `
    -DisplayOutputType StandardOutput

제한 사항

Linux 기반 HDInsight에에 대한 제한 사항은 다음과 같습니다.

  • 대량 내보내기: SQL로 데이터를 내보내는 데 사용되는 Sqoop 커넥터는 현재 대량 삽입을 지원하지 않습니다.

  • 일괄 처리 - 삽입을 수행할 때 -batch 스위치를 사용하면 Sqoop에서 삽입 작업을 일괄 처리하는 대신 여러 번의 삽입 작업을 수행합니다.

다음 단계

이제 Sqoop을 사용하는 방법에 대해 알아봤습니다. 자세한 내용은 다음을 참조하세요.