You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-commits@jackrabbit.apache.org by am...@apache.org on 2015/09/11 09:28:27 UTC
svn commit: r1702387 - in /jackrabbit/oak/trunk/oak-core/src:
main/java/org/apache/jackrabbit/oak/plugins/blob/
main/java/org/apache/jackrabbit/oak/plugins/blob/datastore/
test/java/org/apache/jackrabbit/oak/plugins/blob/
test/java/org/apache/jackrabbi...
Author: amitj
Date: Fri Sep 11 07:28:27 2015
New Revision: 1702387
URL: http://svn.apache.org/r1702387
Log:
OAK-3360: Tracking the start time of mark in GC for a shared datastore
Changes made to track the start time by by creating a start time marker before starting reference collection and clearing the files after sweep
Modified:
jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/BlobGC.java
jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/GarbageCollectionRepoStats.java
jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/MarkSweepGarbageCollector.java
jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/datastore/SharedDataStoreUtils.java
jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/blob/SharedDataStoreUtilsTest.java
jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/document/SharedBlobStoreGCTest.java
Modified: jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/BlobGC.java
URL: http://svn.apache.org/viewvc/jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/BlobGC.java?rev=1702387&r1=1702386&r2=1702387&view=diff
==============================================================================
--- jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/BlobGC.java (original)
+++ jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/BlobGC.java Fri Sep 11 07:28:27 2015
@@ -138,7 +138,8 @@ public class BlobGC extends AnnotatedSt
private CompositeDataSupport toCompositeData(GarbageCollectionRepoStats statObj) throws OpenDataException {
Object[] values = new Object[] {
statObj.getRepositoryId(),
- (statObj.getLastModified() == 0 ? "" : (new Date(statObj.getLastModified()))).toString(),
+ (statObj.getStartTime() == 0 ? "" : (new Date(statObj.getStartTime()))).toString(),
+ (statObj.getEndTime() == 0 ? "" : (new Date(statObj.getEndTime()))).toString(),
statObj.getLength(),
humanReadableByteCount(statObj.getLength()),
statObj.getNumLines()
@@ -148,15 +149,17 @@ public class BlobGC extends AnnotatedSt
private static final String[] FIELD_NAMES = new String[] {
"repositoryId",
- "referencesLastModifiedTime",
+ "markStartTime",
+ "markEndTime",
"referenceFileSizeBytes",
"referencesFileSize",
"numReferences",
};
private static final String[] FIELD_DESCRIPTIONS = new String[] {
- "Repository ID",
- "Last modified time of references",
+ "Repository ID",
+ "Start time of mark",
+ "End time of mark",
"References file size in bytes",
"References file size in human readable format",
"Number of references"
@@ -165,6 +168,7 @@ public class BlobGC extends AnnotatedSt
private static final OpenType[] FIELD_TYPES = new OpenType[] {
SimpleType.STRING,
SimpleType.STRING,
+ SimpleType.STRING,
SimpleType.LONG,
SimpleType.STRING,
SimpleType.INTEGER
Modified: jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/GarbageCollectionRepoStats.java
URL: http://svn.apache.org/viewvc/jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/GarbageCollectionRepoStats.java?rev=1702387&r1=1702386&r2=1702387&view=diff
==============================================================================
--- jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/GarbageCollectionRepoStats.java (original)
+++ jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/GarbageCollectionRepoStats.java Fri Sep 11 07:28:27 2015
@@ -22,7 +22,9 @@ package org.apache.jackrabbit.oak.plugin
public class GarbageCollectionRepoStats {
private String repositoryId;
- private long lastModified;
+ private long startTime;
+
+ private long endTime;
private long length;
@@ -36,12 +38,12 @@ public class GarbageCollectionRepoStats
this.repositoryId = repositoryId;
}
- public long getLastModified() {
- return lastModified;
+ public long getEndTime() {
+ return endTime;
}
- public void setLastModified(long lastModified) {
- this.lastModified = lastModified;
+ public void setEndTime(long endTime) {
+ this.endTime = endTime;
}
public long getLength() {
@@ -59,4 +61,12 @@ public class GarbageCollectionRepoStats
public int getNumLines() {
return numLines;
}
+
+ public long getStartTime() {
+ return startTime;
+ }
+
+ public void setStartTime(long startTime) {
+ this.startTime = startTime;
+ }
}
Modified: jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/MarkSweepGarbageCollector.java
URL: http://svn.apache.org/viewvc/jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/MarkSweepGarbageCollector.java?rev=1702387&r1=1702386&r2=1702387&view=diff
==============================================================================
--- jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/MarkSweepGarbageCollector.java (original)
+++ jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/MarkSweepGarbageCollector.java Fri Sep 11 07:28:27 2015
@@ -17,6 +17,7 @@
package org.apache.jackrabbit.oak.plugins.blob;
import java.io.BufferedWriter;
+import java.io.ByteArrayInputStream;
import java.io.Closeable;
import java.io.File;
import java.io.FileInputStream;
@@ -183,6 +184,16 @@ public class MarkSweepGarbageCollector i
}
});
+ // Get all the markers available
+ List<DataRecord> markerFiles =
+ ((SharedDataStore) blobStore).getAllMetadataRecords(SharedStoreRecordType.MARKED_START_MARKER.getType());
+ Map<String, DataRecord> markers = Maps.uniqueIndex(markerFiles, new Function<DataRecord, String>() {
+ @Override
+ public String apply(DataRecord input) {
+ return SharedStoreRecordType.MARKED_START_MARKER.getIdFromName(input.getIdentifier().toString());
+ }
+ });
+
// Get all the repositories registered
List<DataRecord> repoFiles =
((SharedDataStore) blobStore).getAllMetadataRecords(SharedStoreRecordType.REPOSITORY.getType());
@@ -193,9 +204,13 @@ public class MarkSweepGarbageCollector i
stat.setRepositoryId(repoId);
if (references.containsKey(repoId)) {
DataRecord refRec = references.get(repoId);
- stat.setLastModified(refRec.getLastModified());
+ stat.setEndTime(refRec.getLastModified());
stat.setLength(refRec.getLength());
-
+
+ if (markers.containsKey(repoId)) {
+ stat.setStartTime(markers.get(repoId).getLastModified());
+ }
+
LineNumberReader reader = null;
try {
reader = new LineNumberReader(new InputStreamReader(refRec.getStream()));
@@ -247,7 +262,10 @@ public class MarkSweepGarbageCollector i
*/
private void mark(GarbageCollectorFileState fs) throws IOException, DataStoreException {
LOG.debug("Starting mark phase of the garbage collector");
-
+
+ // Create a time marker in the data store if applicable
+ GarbageCollectionType.get(blobStore).addMarkedStartMarker(blobStore, repoId);
+
// Mark all used references
iterateNodeTree(fs);
@@ -696,13 +714,14 @@ public class MarkSweepGarbageCollector i
enum GarbageCollectionType {
SHARED {
/**
- * Remove the maked references from the blob store root. Default NOOP.
+ * Remove the maked references and the marked markers from the blob store root. Default NOOP.
*
* @param blobStore the blobStore instance
*/
@Override
void removeAllMarkedReferences(GarbageCollectableBlobStore blobStore) {
((SharedDataStore) blobStore).deleteAllMetadataRecords(SharedStoreRecordType.REFERENCES.getType());
+ ((SharedDataStore) blobStore).deleteAllMetadataRecords(SharedStoreRecordType.MARKED_START_MARKER.getType());
}
/**
@@ -738,8 +757,18 @@ public class MarkSweepGarbageCollector i
}
GarbageCollectorFileState.merge(files, fs.getMarkedRefs());
+
+ // Get the timestamp to indicate the earliest mark phase start
+ List<DataRecord> markerFiles =
+ ((SharedDataStore) blobStore).getAllMetadataRecords(
+ SharedStoreRecordType.MARKED_START_MARKER.getType());
+ long earliestMarker = SharedDataStoreUtils.getEarliestRecord(markerFiles).getLastModified();
+ LOG.trace("Earliest marker timestamp {}", earliestMarker);
+
+ long earliestRef = SharedDataStoreUtils.getEarliestRecord(refFiles).getLastModified();
+ LOG.trace("Earliest ref timestamp {}", earliestRef);
- return SharedDataStoreUtils.getEarliestRecord(refFiles).getLastModified();
+ return (earliestMarker < earliestRef ? earliestMarker : earliestRef);
} else {
LOG.error("Not all repositories have marked references available : {}", unAvailRepos);
throw new IOException("Not all repositories have marked references available");
@@ -766,6 +795,17 @@ public class MarkSweepGarbageCollector i
Closeables.close(is, false);
}
}
+
+ @Override
+ public void addMarkedStartMarker(GarbageCollectableBlobStore blobStore, String repoId) {
+ try {
+ ((SharedDataStore) blobStore).addMetadataRecord(new ByteArrayInputStream(new byte[0]),
+ SharedStoreRecordType.MARKED_START_MARKER
+ .getNameFromId(repoId));
+ } catch (DataStoreException e) {
+ LOG.debug("Error creating marked time marker for repo : {}", repoId);
+ }
+ }
},
DEFAULT;
@@ -790,5 +830,7 @@ public class MarkSweepGarbageCollector i
}
return DEFAULT;
}
+
+ public void addMarkedStartMarker(GarbageCollectableBlobStore blobStore, String repoId) {}
}
}
Modified: jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/datastore/SharedDataStoreUtils.java
URL: http://svn.apache.org/viewvc/jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/datastore/SharedDataStoreUtils.java?rev=1702387&r1=1702386&r2=1702387&view=diff
==============================================================================
--- jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/datastore/SharedDataStoreUtils.java (original)
+++ jackrabbit/oak/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/datastore/SharedDataStoreUtils.java Fri Sep 11 07:28:27 2015
@@ -96,7 +96,7 @@ public class SharedDataStoreUtils {
* Encapsulates the different type of records at the data store root.
*/
public enum SharedStoreRecordType {
- REFERENCES("references"), REPOSITORY("repository");
+ REFERENCES("references"), REPOSITORY("repository"), MARKED_START_MARKER("markedTimestamp");
private final String type;
Modified: jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/blob/SharedDataStoreUtilsTest.java
URL: http://svn.apache.org/viewvc/jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/blob/SharedDataStoreUtilsTest.java?rev=1702387&r1=1702386&r2=1702387&view=diff
==============================================================================
--- jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/blob/SharedDataStoreUtilsTest.java (original)
+++ jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/blob/SharedDataStoreUtilsTest.java Fri Sep 11 07:28:27 2015
@@ -67,8 +67,18 @@ public class SharedDataStoreUtilsTest {
dataStore.addMetadataRecord(new ByteArrayInputStream(new byte[0]),
SharedStoreRecordType.REPOSITORY.getNameFromId(repoId2));
DataRecord repo2 = dataStore.getMetadataRecord(SharedStoreRecordType.REPOSITORY.getNameFromId(repoId2));
-
- // Add reference records
+
+ // Add reference marker record for repo1
+ dataStore.addMetadataRecord(new ByteArrayInputStream(new byte[0]),
+ SharedStoreRecordType.MARKED_START_MARKER.getNameFromId(repoId1));
+ DataRecord markerRec1 = dataStore.getMetadataRecord(SharedStoreRecordType.MARKED_START_MARKER.getNameFromId(repoId1));
+ Assert.assertEquals(
+ SharedStoreRecordType.MARKED_START_MARKER.getIdFromName(markerRec1.getIdentifier().toString()),
+ repoId1);
+ long lastModifiedMarkerRec1 = markerRec1.getLastModified();
+ TimeUnit.MILLISECONDS.sleep(100);
+
+ // Add reference records
dataStore.addMetadataRecord(new ByteArrayInputStream(new byte[0]),
SharedStoreRecordType.REFERENCES.getNameFromId(repoId1));
DataRecord rec1 = dataStore.getMetadataRecord(SharedStoreRecordType.REFERENCES.getNameFromId(repoId1));
@@ -97,14 +107,27 @@ public class SharedDataStoreUtilsTest {
// Since, we don't care about which file specifically but only the earliest timestamped record
// Earliest time should be the min timestamp from the 2 reference files
- Assert.assertEquals(SharedDataStoreUtils
- .getEarliestRecord(dataStore.getAllMetadataRecords(SharedStoreRecordType.REFERENCES.getType()))
- .getLastModified(), (lastModifiedRec1 <= lastModifiedRec2 ? lastModifiedRec1 : lastModifiedRec2));
-
+ long minRefTime = (lastModifiedRec1 <= lastModifiedRec2 ? lastModifiedRec1 : lastModifiedRec2);
+ Assert.assertEquals(
+ SharedDataStoreUtils.getEarliestRecord(
+ dataStore.getAllMetadataRecords(SharedStoreRecordType.REFERENCES.getType())).getLastModified(),
+ minRefTime);
+
+ // the marker timestamp should be the minimum
+ long minMarkerTime =
+ SharedDataStoreUtils.getEarliestRecord(
+ dataStore.getAllMetadataRecords(SharedStoreRecordType.MARKED_START_MARKER.getType()))
+ .getLastModified();
+ Assert.assertTrue(minRefTime > minMarkerTime);
+
// Delete references and check back if deleted
dataStore.deleteAllMetadataRecords(SharedStoreRecordType.REFERENCES.getType());
Assert.assertTrue(dataStore.getAllMetadataRecords(SharedStoreRecordType.REFERENCES.getType()).isEmpty());
-
+
+ // Delete markers and check back if deleted
+ dataStore.deleteAllMetadataRecords(SharedStoreRecordType.MARKED_START_MARKER.getType());
+ Assert.assertTrue(dataStore.getAllMetadataRecords(SharedStoreRecordType.MARKED_START_MARKER.getType()).isEmpty());
+
// Repository ids should still be available
Assert.assertEquals(2,
dataStore.getAllMetadataRecords(SharedStoreRecordType.REPOSITORY.getType()).size());
Modified: jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/document/SharedBlobStoreGCTest.java
URL: http://svn.apache.org/viewvc/jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/document/SharedBlobStoreGCTest.java?rev=1702387&r1=1702386&r2=1702387&view=diff
==============================================================================
--- jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/document/SharedBlobStoreGCTest.java (original)
+++ jackrabbit/oak/trunk/oak-core/src/test/java/org/apache/jackrabbit/oak/plugins/document/SharedBlobStoreGCTest.java Fri Sep 11 07:28:27 2015
@@ -155,6 +155,7 @@ public class SharedBlobStoreGCTest {
for (GarbageCollectionRepoStats stat : statsList) {
observedNumBlobs.add(stat.getNumLines());
observedRepoIds.add(stat.getRepositoryId());
+ Assert.assertTrue(stat.getStartTime() <= stat.getEndTime());
}
Assert.assertTrue(Sets.difference(actualNumBlobs, observedNumBlobs).isEmpty());