Joerg Hoh created OAK-10116:
-------------------------------
Summary: Performance problem when importing nodes with many binary
properties and remote blobstore
Key: OAK-10116
URL: https://issues.apache.org/jira/browse/OAK-10116
Project: Jackrabbit Oak
Issue Type: Improvement
Components: blob-cloud, blob-plugins, jcr
Affects Versions: 1.48.0
Reporter: Joerg Hoh
We often import binaryless packages (using JR filevault) into our Oak
instances, which are using a remote blobstore.
We observe bad performance when we import nodes with binary properties. In this
case stacktraces often look like this:
{noformat}
"Queue Processor for Subscriber agent publishSubscriber" #311 daemon prio=5
os_prio=0 cpu=298928.76ms elapsed=576.04s tid=0x0000563f968c6800 nid=0x1644
runnable [0x00007f2a609e3000]
java.lang.Thread.State: RUNNABLE
at java.net.SocketInputStream.socketRead0([email protected]/Native
Method)
at
java.net.SocketInputStream.socketRead([email protected]/SocketInputStream.java:115)
at
java.net.SocketInputStream.read([email protected]/SocketInputStream.java:168)
at
java.net.SocketInputStream.read([email protected]/SocketInputStream.java:140)
at
sun.security.ssl.SSLSocketInputRecord.read([email protected]/SSLSocketInputRecord.java:478)
at
sun.security.ssl.SSLSocketInputRecord.readHeader([email protected]/SSLSocketInputRecord.java:472)
at
sun.security.ssl.SSLSocketInputRecord.bytesInCompletePacket([email protected]/SSLSocketInputRecord.java:70)
at
sun.security.ssl.SSLSocketImpl.readApplicationRecord([email protected]/SSLSocketImpl.java:1328)
at
sun.security.ssl.SSLSocketImpl$AppInputStream.read([email protected]/SSLSocketImpl.java:971)
at
java.io.BufferedInputStream.fill([email protected]/BufferedInputStream.java:252)
at
java.io.BufferedInputStream.read1([email protected]/BufferedInputStream.java:292)
at
java.io.BufferedInputStream.read([email protected]/BufferedInputStream.java:351)
- locked <0x00000007d98d0ca8> (a java.io.BufferedInputStream)
at
sun.net.www.http.HttpClient.parseHTTPHeader([email protected]/HttpClient.java:746)
at
sun.net.www.http.HttpClient.parseHTTP([email protected]/HttpClient.java:689)
at
sun.net.www.protocol.http.HttpURLConnection.getInputStream0([email protected]/HttpURLConnection.java:1615)
- locked <0x00000007d98cb480> (a
sun.net.www.protocol.https.DelegateHttpsURLConnection)
at
sun.net.www.protocol.http.HttpURLConnection.getInputStream([email protected]/HttpURLConnection.java:1520)
- locked <0x00000007d98cb480> (a
sun.net.www.protocol.https.DelegateHttpsURLConnection)
at
java.net.HttpURLConnection.getResponseCode([email protected]/HttpURLConnection.java:527)
at
sun.net.www.protocol.https.HttpsURLConnectionImpl.getResponseCode([email protected]/HttpsURLConnectionImpl.java:334)
at
com.microsoft.azure.storage.core.ExecutionEngine.executeWithRetry(ExecutionEngine.java:115)
at
com.microsoft.azure.storage.blob.CloudBlob.downloadAttributes(CloudBlob.java:1414)
at
com.microsoft.azure.storage.blob.CloudBlob.downloadAttributes(CloudBlob.java:1381)
at
org.apache.jackrabbit.oak.blob.cloud.azure.blobstorage.AzureBlobStoreBackend.getRecord(AzureBlobStoreBackend.java:408)
at
org.apache.jackrabbit.oak.plugins.blob.AbstractSharedCachingDataStore.getRecordIfStored(AbstractSharedCachingDataStore.java:210)
at
org.apache.jackrabbit.core.data.AbstractDataStore.getRecordFromReference(AbstractDataStore.java:72)
at
org.apache.jackrabbit.oak.plugins.blob.datastore.DataStoreBlobStore.getBlobId(DataStoreBlobStore.java:402)
at
org.apache.jackrabbit.oak.segment.SegmentNodeStore.getBlob(SegmentNodeStore.java:257)
at
org.apache.jackrabbit.oak.composite.CompositeNodeStore.getBlob(CompositeNodeStore.java:202)
at
org.apache.jackrabbit.oak.core.MutableRoot.getBlob(MutableRoot.java:342)
at
org.apache.jackrabbit.oak.plugins.value.jcr.ValueFactoryImpl.createValue(ValueFactoryImpl.java:111)
at
org.apache.jackrabbit.vault.util.DocViewProperty.apply(DocViewProperty.java:413)
at
org.apache.jackrabbit.vault.fs.impl.io.DocViewSAXImporter.createNode(DocViewSAXImporter.java:1131)
at
org.apache.jackrabbit.vault.fs.impl.io.DocViewSAXImporter.addNode(DocViewSAXImporter.java:891)
at
org.apache.jackrabbit.vault.fs.impl.io.DocViewSAXImporter.startElement(DocViewSAXImporter.java:681)
at
com.sun.org.apache.xerces.internal.parsers.AbstractSAXParser.startElement([email protected]/AbstractSAXParser.java:510)
at
com.sun.org.apache.xerces.internal.impl.XMLNSDocumentScannerImpl.scanStartElement([email protected]/XMLNSDocumentScannerImpl.java:374)
at
com.sun.org.apache.xerces.internal.impl.XMLDocumentFragmentScannerImpl$FragmentContentDriver.next([email protected]/XMLDocumentFragmentScannerImpl.java:2710)
at
com.sun.org.apache.xerces.internal.impl.XMLDocumentScannerImpl.next([email protected]/XMLDocumentScannerImpl.java:605)
at
com.sun.org.apache.xerces.internal.impl.XMLNSDocumentScannerImpl.next([email protected]/XMLNSDocumentScannerImpl.java:112)
at
com.sun.org.apache.xerces.internal.impl.XMLDocumentFragmentScannerImpl.scanDocument([email protected]/XMLDocumentFragmentScannerImpl.java:534)
at
com.sun.org.apache.xerces.internal.parsers.XML11Configuration.parse([email protected]/XML11Configuration.java:888)
at
com.sun.org.apache.xerces.internal.parsers.XML11Configuration.parse([email protected]/XML11Configuration.java:824)
at
com.sun.org.apache.xerces.internal.parsers.XMLParser.parse([email protected]/XMLParser.java:141)
at
com.sun.org.apache.xerces.internal.parsers.AbstractSAXParser.parse([email protected]/AbstractSAXParser.java:1216)
at
com.sun.org.apache.xerces.internal.jaxp.SAXParserImpl$JAXPSAXParser.parse([email protected]/SAXParserImpl.java:635)
at
com.sun.org.apache.xerces.internal.jaxp.SAXParserImpl.parse([email protected]/SAXParserImpl.java:324)
at
org.apache.jackrabbit.vault.fs.impl.io.GenericArtifactHandler.accept(GenericArtifactHandler.java:100)
at org.apache.jackrabbit.vault.fs.io.Importer.commit(Importer.java:932)
at org.apache.jackrabbit.vault.fs.io.Importer.commit(Importer.java:799)
{noformat}
In this context we can ensure that all binaries are available on the remote
blobstore, so a call to the blobstore would not required, at least not for
validating its presence; all other information could/should be part of the
filevault package.
In my opinion the ValueFactory should be able to create a binary property
without reaching out to the blobstore to avoid the network latency. This would
speed up the import process dramatically; as in the context of this situation
we can create approx 20 binary properties per second, while we can create
thousands of non-binary properties in the same time.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)