[jira] [Updated] (OAK-4565) S3Backend fails to upload large metadata records

["Julian Reschke (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/OAK-4565?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Julian Reschke updated OAK-4565:
--------------------------------
    Fix Version/s: 1.6

> S3Backend fails to upload large metadata records
> ------------------------------------------------
>
>                 Key: OAK-4565
>                 URL: https://issues.apache.org/jira/browse/OAK-4565
>             Project: Jackrabbit Oak
>          Issue Type: Bug
>          Components: blob
>    Affects Versions: 1.4.5
>            Reporter: Martin Wehner
>            Assignee: Amit Jain
>              Labels: gc, s3
>             Fix For: 1.6, 1.5.6, 1.4.6, 1.2.18
>
>         Attachments: OAK-4565-1.4.patch, OAK-4565-TRUNK.patch
>
>
> If a large enough metadata record is added to a S3 DS (like the list of blob references collected during the mark phase of the MarkSweepGC) the upload will fail (i.e. never start). This is caused by {{S3Backend.addMetadataRecord()}} providing an InputStream to the S3 TransferManager without specifying the size in the Metadata. 
> A warning to this effect is logged by the AWS SDK each time you add a metadata record: 
> {noformat}
> [s3-transfer-manager-worker-1] AmazonS3Client.java:1364 No content length specified for stream data.  Stream contents will be buffered in memory and could result in out of memory errors.
> {noformat}
> Normally this shouldn't be too big of a problem but in a repository with over 36 million blob references the list of marked refs produced by the GC is over 5GB. In this case the S3 transfer worker thread will be stuck in a seemingly endless loop where it tries to allocate the memory reading the file into memory and never finishes (although the JVM has 80GB of heap), eating away resources in the process:
> {noformat}
>    java.lang.Thread.State: RUNNABLE
> 	at org.apache.http.util.ByteArrayBuffer.append(ByteArrayBuffer.java:90)
> 	at org.apache.http.util.EntityUtils.toByteArray(EntityUtils.java:137)
> 	at org.apache.http.entity.BufferedHttpEntity.<init>(BufferedHttpEntity.java:63)
> 	at com.amazonaws.http.HttpRequestFactory.newBufferedHttpEntity(HttpRequestFactory.java:247)
> 	at com.amazonaws.http.HttpRequestFactory.createHttpRequest(HttpRequestFactory.java:126)
> 	at com.amazonaws.http.AmazonHttpClient$ExecOneRequestParams.newApacheRequest(AmazonHttpClient.java:650)
> 	at com.amazonaws.http.AmazonHttpClient.executeOneRequest(AmazonHttpClient.java:730)
> 	at com.amazonaws.http.AmazonHttpClient.executeHelper(AmazonHttpClient.java:505)
> 	at com.amazonaws.http.AmazonHttpClient.execute(AmazonHttpClient.java:317)
> 	at com.amazonaws.services.s3.AmazonS3Client.invoke(AmazonS3Client.java:3595)
> 	at com.amazonaws.services.s3.AmazonS3Client.putObject(AmazonS3Client.java:1382)
> 	at com.amazonaws.services.s3.transfer.internal.UploadCallable.uploadInOneChunk(UploadCallable.java:131)
> 	at com.amazonaws.services.s3.transfer.internal.UploadCallable.call(UploadCallable.java:123)
> 	at com.amazonaws.services.s3.transfer.internal.UploadMonitor.call(UploadMonitor.java:139)
> 	at com.amazonaws.services.s3.transfer.internal.UploadMonitor.call(UploadMonitor.java:47)
> 	at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> 	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
> 	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
> 	at java.lang.Thread.run(Thread.java:745)
> {noformat} 
> The last log message by the GC thread will be like this:
> {noformat}
> *INFO* [sling-oak-observation-1273] org.apache.jackrabbit.oak.plugins.blob.MarkSweepGarbageCollector Number of valid blob references marked under mark phase of Blob garbage collection [36147734]
> {noformat} 
> followed by the above AWS warning, then it will stall waiting for the transfer to finish.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)