hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Steve Loughran (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HADOOP-14700) NativeAzureFileSystem.open() ignores blob container name
Date Tue, 01 Aug 2017 21:33:00 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-14700?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16109826#comment-16109826
] 

Steve Loughran commented on HADOOP-14700:
-----------------------------------------

looks to me like pathB is set to a path in container A, at least in that demo code {{new Path(s"wasbs://$CONTAINER_A@$ACCOUNT.blob.core.windows.net/$TEST_BLOB")}}.

Does it still happen if you change it.

If it does, it may be that the CONTAINER_A value isn't being used as part of the URL used
to cache FS instances, so the first container instance used is handed back

> NativeAzureFileSystem.open() ignores blob container name
> --------------------------------------------------------
>
>                 Key: HADOOP-14700
>                 URL: https://issues.apache.org/jira/browse/HADOOP-14700
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: fs
>    Affects Versions: 3.0.0-beta1, 3.0.0-alpha4
>            Reporter: Cheng Lian
>
> {{NativeAzureFileSystem}} instances are associated with the blob container used to initialize
the file system. Assuming that a file system instance {{fs}} is associated with a container
{{A}}, when trying to access a blob inside another container {{B}}, {{fs}} still tries to
find the blob inside container {{A}}. If there happens to be two blobs with the same name
inside both containers, the user may get a wrong result because {{fs}} reads the contents
from the blob inside container {{A}} instead of container {{B}}.
> The following self-contained Scala code snippet illustrates this issue. You may reproduce
it by running the script inside the [Ammonite REPL|http://ammonite.io/].
> {code}
> import $ivy.`com.jsuereth::scala-arm:2.0`
> import $ivy.`com.microsoft.azure:azure-storage:5.2.0`
> import $ivy.`org.apache.hadoop:hadoop-azure:3.0.0-alpha4`
> import $ivy.`org.apache.hadoop:hadoop-common:3.0.0-alpha4`
> import $ivy.`org.scalatest::scalatest:3.0.3`
> import java.io.{BufferedReader, InputStreamReader}
> import java.net.URI
> import java.time.{Duration, Instant}
> import java.util.{Date, EnumSet}
> import com.microsoft.azure.storage.{CloudStorageAccount, StorageCredentialsAccountAndKey}
> import com.microsoft.azure.storage.blob.{SharedAccessBlobPermissions, SharedAccessBlobPolicy}
> import org.apache.hadoop.conf.Configuration
> import org.apache.hadoop.fs.{FileSystem, Path}
> import org.apache.hadoop.fs.azure.{AzureException, NativeAzureFileSystem}
> import org.scalatest.Assertions._
> import resource._
> // Utility implicit conversion for auto resource management.
> implicit def `Closable->Resource`[T <: { def close() }]: Resource[T] = new Resource[T]
{
>   override def close(closable: T): Unit = closable.close()
> }
> // Credentials information
> val ACCOUNT = "** REDACTED **"
> val ACCESS_KEY = "** REDACTED **"
> // We'll create two different containers, both contain a blob named "test-blob" but with
different
> // contents.
> val CONTAINER_A = "container-a"
> val CONTAINER_B = "container-b"
> val TEST_BLOB = "test-blob"
> val blobClient = {
>   val credentials = new StorageCredentialsAccountAndKey(ACCOUNT, ACCESS_KEY)
>   val account = new CloudStorageAccount(credentials, /* useHttps */ true)
>   account.createCloudBlobClient()
> }
> // Generates a read-only SAS key restricted within "container-a".
> val sasKeyForContainerA = {
>   val since = Instant.now() minus Duration.ofMinutes(10)
>   val duration = Duration.ofHours(1)
>   val policy = new SharedAccessBlobPolicy()
>   policy.setSharedAccessStartTime(Date.from(since))
>   policy.setSharedAccessExpiryTime(Date.from(since plus duration))
>   policy.setPermissions(EnumSet.of(
>     SharedAccessBlobPermissions.READ,
>     SharedAccessBlobPermissions.LIST
>   ))
>   blobClient
>     .getContainerReference(CONTAINER_A)
>     .generateSharedAccessSignature(policy, null)
> }
> // Sets up testing containers and blobs using the Azure storage SDK:
> //
> //   container-a/test-blob => "foo"
> //   container-b/test-blob => "bar"
> {
>   val containerARef = blobClient.getContainerReference(CONTAINER_A)
>   val containerBRef = blobClient.getContainerReference(CONTAINER_B)
>   containerARef.createIfNotExists()
>   containerARef.getBlockBlobReference(TEST_BLOB).uploadText("foo")
>   containerBRef.createIfNotExists()
>   containerBRef.getBlockBlobReference(TEST_BLOB).uploadText("bar")
> }
> val pathA = new Path(s"wasbs://$CONTAINER_A@$ACCOUNT.blob.core.windows.net/$TEST_BLOB")
> val pathB = new Path(s"wasbs://$CONTAINER_A@$ACCOUNT.blob.core.windows.net/$TEST_BLOB")
> for {
>   // Creates a file system associated with "container-a".
>   fs <- managed {
>     val conf = new Configuration
>     conf.set("fs.wasbs.impl", classOf[NativeAzureFileSystem].getName)
>     conf.set(s"fs.azure.sas.$CONTAINER_A.$ACCOUNT.blob.core.windows.net", sasKeyForContainerA)
>     pathA.getFileSystem(conf)
>   }
>   // Opens a reader pointing to "container-a/test-blob". We expect to get the string
"foo" written
>   // to this blob previously.
>   readerA <- managed(new BufferedReader(new InputStreamReader(fs open pathA)))
>   // Opens a reader pointing to "container-b/test-blob". We expect to get an exception
since the SAS
>   // key used to create the `FileSystem` instance is restricted to "container-a".
>   readerB <- managed(new BufferedReader(new InputStreamReader(fs open pathB)))
> } {
>   // Should get "foo"
>   assert(readerA.readLine() == "foo")
>   // Should catch an exception ...
>   intercept[AzureException] {
>     // ... but instead, we get string "foo" here, which indicates that the readerB was
reading from
>     // "container-a" instead of "container-b".
>     val contents = readerB.readLine()
>     println(s"Should not reach here but we got $contents")
>   }
> }
> {code}



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org


Mime
View raw message