swamirishi commented on code in PR #9079:
URL: https://github.com/apache/ozone/pull/9079#discussion_r2413442520


##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1223,6 +1273,81 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
             toSnapshot.getSnapshotTableKey()));
   }
 
+  /**
+   * Resolve the VersionMeta of the ancestor snapshot (fromSnapshotId)
+   * that the given snapshot (toSnapshot) was built on.
+   * Traverses the snapshot chain backwards using prevSnapId.
+   *
+   * @param toSnapshot the target snapshot.
+   * @param fromSnapshotId UUID of the source snapshot.
+   * @return the resolved VersionMeta of the source that was used to build the 
target.
+   */
+  private OmSnapshotLocalData.VersionMeta resolveBaseVersionMeta(
+      OmSnapshotLocalData toSnapshot,
+      UUID fromSnapshotId) throws IOException {
+    OmMetadataManagerImpl metadataManager =
+        (OmMetadataManagerImpl) ozoneManager.getMetadataManager();
+    // Start walking back from the child snapshot
+    OmSnapshotLocalData child = toSnapshot;
+    int numSnapshotsTraversed = 0;
+    while (numSnapshotsTraversed <= maxSnapshotLimit && 
!child.getPreviousSnapshotId().equals(fromSnapshotId)) {
+      UUID parentId = child.getPreviousSnapshotId();
+      // Load the parent snapshot in the chain
+      child = getSnapshotLocalData(
+          getSnapshotInfo(ozoneManager,
+              metadataManager.getSnapshotChainManager(),
+              parentId));
+      numSnapshotsTraversed++;
+    }
+    if (numSnapshotsTraversed > maxSnapshotLimit) {
+      LOG.error("Exceeded the traversal limit of {} while finding the 
VersionMeta of the fromSnapshot : {}" +
+          " that the toSnapshot was built on.", maxSnapshotLimit, 
fromSnapshotId);
+      return null;
+    }
+    SnapshotInfo snapshotInfo =
+        getSnapshotInfo(ozoneManager, 
metadataManager.getSnapshotChainManager(), fromSnapshotId);
+    OmSnapshotLocalData fromSnapshot = getSnapshotLocalData(snapshotInfo);
+    // Get the version that the child was built from
+    OmSnapshotLocalData.VersionMeta childVersionMeta = 
child.getVersionSstFileInfos().get(child.getVersion());

Review Comment:
   This logic of finding childVersionMeta is wrong. We need to update the 
version inside the loop.
   



##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1180,6 +1199,37 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
     // TODO: [SNAPSHOT] Refactor the parameter list
     Optional<Set<String>> deltaFiles = Optional.empty();
 
+    OmSnapshotLocalData fromSnapshotLocalData = getSnapshotLocalData(fsInfo);
+    int fromSnapshotVersion = fromSnapshotLocalData.getVersion();
+    OmSnapshotLocalData toSnapshotLocalData = getSnapshotLocalData(tsInfo);
+    int toSnapshotVersion = toSnapshotLocalData.getVersion();
+
+    try {
+      if (fromSnapshotVersion > 0 && toSnapshotVersion > 0) {

Review Comment:
   We should not check both. Just checking toSnapshotVersion should be good 
enough as the defrag service would always run from the beginning of the chain



##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1180,6 +1199,37 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
     // TODO: [SNAPSHOT] Refactor the parameter list
     Optional<Set<String>> deltaFiles = Optional.empty();
 
+    OmSnapshotLocalData fromSnapshotLocalData = getSnapshotLocalData(fsInfo);
+    int fromSnapshotVersion = fromSnapshotLocalData.getVersion();
+    OmSnapshotLocalData toSnapshotLocalData = getSnapshotLocalData(tsInfo);
+    int toSnapshotVersion = toSnapshotLocalData.getVersion();
+
+    try {
+      if (fromSnapshotVersion > 0 && toSnapshotVersion > 0) {
+        // both snapshots are defragmented, To calculate snap-diff, we can 
simply compare the
+        // SST files contained in OmSnapshotLocalData instances of both these 
and get the delta files
+        OmSnapshotLocalData.VersionMeta toSnapVersionMeta =
+            
toSnapshotLocalData.getVersionSstFileInfos().get(toSnapshotVersion);

Review Comment:
   Let's create a function for this in OmSnapshotLocalData



##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1223,6 +1273,81 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
             toSnapshot.getSnapshotTableKey()));
   }
 
+  /**
+   * Resolve the VersionMeta of the ancestor snapshot (fromSnapshotId)
+   * that the given snapshot (toSnapshot) was built on.
+   * Traverses the snapshot chain backwards using prevSnapId.
+   *
+   * @param toSnapshot the target snapshot.
+   * @param fromSnapshotId UUID of the source snapshot.
+   * @return the resolved VersionMeta of the source that was used to build the 
target.
+   */
+  private OmSnapshotLocalData.VersionMeta resolveBaseVersionMeta(
+      OmSnapshotLocalData toSnapshot,
+      UUID fromSnapshotId) throws IOException {
+    OmMetadataManagerImpl metadataManager =
+        (OmMetadataManagerImpl) ozoneManager.getMetadataManager();
+    // Start walking back from the child snapshot
+    OmSnapshotLocalData child = toSnapshot;
+    int numSnapshotsTraversed = 0;
+    while (numSnapshotsTraversed <= maxSnapshotLimit && 
!child.getPreviousSnapshotId().equals(fromSnapshotId)) {
+      UUID parentId = child.getPreviousSnapshotId();
+      // Load the parent snapshot in the chain
+      child = getSnapshotLocalData(
+          getSnapshotInfo(ozoneManager,
+              metadataManager.getSnapshotChainManager(),
+              parentId));
+      numSnapshotsTraversed++;

Review Comment:
   ```suggestion
         resolvedChildVersionMeta = 
child.getVersionMeta(resolvedChildVersionMeta.getPreviousSnapshotVersion());
         numSnapshotsTraversed++;
   ```



##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1223,6 +1273,81 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
             toSnapshot.getSnapshotTableKey()));
   }
 
+  /**
+   * Resolve the VersionMeta of the ancestor snapshot (fromSnapshotId)
+   * that the given snapshot (toSnapshot) was built on.
+   * Traverses the snapshot chain backwards using prevSnapId.
+   *
+   * @param toSnapshot the target snapshot.
+   * @param fromSnapshotId UUID of the source snapshot.
+   * @return the resolved VersionMeta of the source that was used to build the 
target.
+   */
+  private OmSnapshotLocalData.VersionMeta resolveBaseVersionMeta(
+      OmSnapshotLocalData toSnapshot,
+      UUID fromSnapshotId) throws IOException {
+    OmMetadataManagerImpl metadataManager =
+        (OmMetadataManagerImpl) ozoneManager.getMetadataManager();
+    // Start walking back from the child snapshot
+    OmSnapshotLocalData child = toSnapshot;
+    int numSnapshotsTraversed = 0;

Review Comment:
   ```suggestion
       int numSnapshotsTraversed = 0;
       VersionMeta resolvedChildVersionMeta = 
child.getVersionMeta(childVersion.getVersion());
   ```



##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1223,6 +1273,81 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
             toSnapshot.getSnapshotTableKey()));
   }
 
+  /**
+   * Resolve the VersionMeta of the ancestor snapshot (fromSnapshotId)
+   * that the given snapshot (toSnapshot) was built on.
+   * Traverses the snapshot chain backwards using prevSnapId.
+   *
+   * @param toSnapshot the target snapshot.
+   * @param fromSnapshotId UUID of the source snapshot.
+   * @return the resolved VersionMeta of the source that was used to build the 
target.
+   */
+  private OmSnapshotLocalData.VersionMeta resolveBaseVersionMeta(
+      OmSnapshotLocalData toSnapshot,
+      UUID fromSnapshotId) throws IOException {
+    OmMetadataManagerImpl metadataManager =
+        (OmMetadataManagerImpl) ozoneManager.getMetadataManager();
+    // Start walking back from the child snapshot
+    OmSnapshotLocalData child = toSnapshot;
+    int numSnapshotsTraversed = 0;
+    while (numSnapshotsTraversed <= maxSnapshotLimit && 
!child.getPreviousSnapshotId().equals(fromSnapshotId)) {
+      UUID parentId = child.getPreviousSnapshotId();
+      // Load the parent snapshot in the chain
+      child = getSnapshotLocalData(
+          getSnapshotInfo(ozoneManager,
+              metadataManager.getSnapshotChainManager(),
+              parentId));
+      numSnapshotsTraversed++;
+    }
+    if (numSnapshotsTraversed > maxSnapshotLimit) {

Review Comment:
   Let us not have this max snapshot limit loop condition. In the worst case we 
are going to support only maximum of 64k snapshots in the chain.



##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1223,6 +1273,81 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
             toSnapshot.getSnapshotTableKey()));
   }
 
+  /**
+   * Resolve the VersionMeta of the ancestor snapshot (fromSnapshotId)
+   * that the given snapshot (toSnapshot) was built on.
+   * Traverses the snapshot chain backwards using prevSnapId.
+   *
+   * @param toSnapshot the target snapshot.
+   * @param fromSnapshotId UUID of the source snapshot.
+   * @return the resolved VersionMeta of the source that was used to build the 
target.
+   */
+  private OmSnapshotLocalData.VersionMeta resolveBaseVersionMeta(
+      OmSnapshotLocalData toSnapshot,
+      UUID fromSnapshotId) throws IOException {
+    OmMetadataManagerImpl metadataManager =
+        (OmMetadataManagerImpl) ozoneManager.getMetadataManager();
+    // Start walking back from the child snapshot
+    OmSnapshotLocalData child = toSnapshot;
+    int numSnapshotsTraversed = 0;
+    while (numSnapshotsTraversed <= maxSnapshotLimit && 
!child.getPreviousSnapshotId().equals(fromSnapshotId)) {
+      UUID parentId = child.getPreviousSnapshotId();
+      // Load the parent snapshot in the chain
+      child = getSnapshotLocalData(
+          getSnapshotInfo(ozoneManager,
+              metadataManager.getSnapshotChainManager(),
+              parentId));
+      numSnapshotsTraversed++;
+    }
+    if (numSnapshotsTraversed > maxSnapshotLimit) {
+      LOG.error("Exceeded the traversal limit of {} while finding the 
VersionMeta of the fromSnapshot : {}" +
+          " that the toSnapshot was built on.", maxSnapshotLimit, 
fromSnapshotId);
+      return null;
+    }
+    SnapshotInfo snapshotInfo =
+        getSnapshotInfo(ozoneManager, 
metadataManager.getSnapshotChainManager(), fromSnapshotId);
+    OmSnapshotLocalData fromSnapshot = getSnapshotLocalData(snapshotInfo);
+    // Get the version that the child was built from
+    OmSnapshotLocalData.VersionMeta childVersionMeta = 
child.getVersionSstFileInfos().get(child.getVersion());
+    int versionUsedWhenBuildingChild = 
childVersionMeta.getPreviousSnapshotVersion();

Review Comment:
   ```suggestion
        return resolvedChildVersionMeta;
   ```



##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1180,6 +1199,37 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
     // TODO: [SNAPSHOT] Refactor the parameter list
     Optional<Set<String>> deltaFiles = Optional.empty();
 
+    OmSnapshotLocalData fromSnapshotLocalData = getSnapshotLocalData(fsInfo);
+    int fromSnapshotVersion = fromSnapshotLocalData.getVersion();
+    OmSnapshotLocalData toSnapshotLocalData = getSnapshotLocalData(tsInfo);
+    int toSnapshotVersion = toSnapshotLocalData.getVersion();
+
+    try {
+      if (fromSnapshotVersion > 0 && toSnapshotVersion > 0) {
+        // both snapshots are defragmented, To calculate snap-diff, we can 
simply compare the
+        // SST files contained in OmSnapshotLocalData instances of both these 
and get the delta files
+        OmSnapshotLocalData.VersionMeta toSnapVersionMeta =
+            
toSnapshotLocalData.getVersionSstFileInfos().get(toSnapshotVersion);
+        // get the source snapshot SST files (versionMeta)  corresponding to 
target snapshot
+        OmSnapshotLocalData.VersionMeta fromSnapVersionMeta =
+            resolveBaseVersionMeta(toSnapshotLocalData, 
fromSnapshot.getSnapshotID());
+        // Calculate diff files using helper method
+        if (toSnapVersionMeta == null) {
+          String errMsg =
+              "Cannot find corresponding version of from snapshot " + 
fromSnapshotVersion + " from " + tsInfo;
+          LOG.error(errMsg);
+          throw new IOException(errMsg);
+        }
+        List<String> diffFiles = calculateDiffFiles(fromSnapVersionMeta, 
toSnapVersionMeta);
+        return OmSnapshotUtils.getSSTDiffListWithFullPath(diffFiles,
+            
OmSnapshotManager.getSnapshotPath(ozoneManager.getMetadataManager(), 
fsInfo).toString(),
+            
OmSnapshotManager.getSnapshotPath(ozoneManager.getMetadataManager(), 
tsInfo).toString(), diffDir);
+      }
+    } catch (Exception e) {
+      LOG.error("Failed to calculate snap-diff between fromSnapshot : {} , 
toSnapshot: {} via optimal method," +
+          "Falling back to other methods", fsInfo, tsInfo, e);
+    }
+
     // Check if compaction DAG is available, use that if so
     if (differ != null && fsInfo != null && tsInfo != null && !useFullDiff) {

Review Comment:
   We need rocksdb checkpoint differ to also use the SnapshotLocalMetadata 
instead of the rocksdb instance itself.
   There are 3 possible conditions:
   
   1. fromSnapshot is defragged and toSNapshot Defragged
   2. fromSnapshot is not defragged and toSnapshot not defragged
   3. fromSnapshot is Defragged and toSnapshot not Defragged
   4. fromSnapshot is not Defragged and toSnapshot Defragged(This is not 
possible.)
   This patch covers only the first condition so far. 2 and 3 condition should 
fall back to dag based diff by using both version 0's sst files to compute diff.
   Effectively logic is :
   ```
   if (toSnapshot.getVersion() > 0) {
      perfromFullDiff(fromSnapshot(version=correspoding version from traversing 
linked list), toSnapshot(version=toSnapshotVersion));
   } else {
    if (Dag available) {
         performDagDiff(fromSnapshot(version=0), toSnapshot(version=0));
   } else {
        peformFullDiff(fromSnapshot(version=0), toSnapshot(version=0))
   }
   }
   }
   
   ```



##########
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/snapshot/SnapshotDiffManager.java:
##########
@@ -1223,6 +1265,70 @@ Set<String> getDeltaFiles(OmSnapshot fromSnapshot,
             toSnapshot.getSnapshotTableKey()));
   }
 
+  /**
+   * Resolve the VersionMeta of the ancestor snapshot (fromSnapshotId)
+   * that the given snapshot (toSnapshot) was built on.
+   *
+   * Traverses the snapshot chain backwards using prevSnapId.
+   */
+  private OmSnapshotLocalData.VersionMeta resolveBaseVersionMeta(
+      OmSnapshotLocalData toSnapshot,
+      UUID fromSnapshotId) throws IOException {
+    OmMetadataManagerImpl metadataManager =
+        (OmMetadataManagerImpl) ozoneManager.getMetadataManager();
+    // Start walking back from the child snapshot
+    OmSnapshotLocalData child = toSnapshot;
+    while (!child.getPreviousSnapshotId().equals(fromSnapshotId)) {
+      UUID parentId = child.getPreviousSnapshotId();
+      // Load the parent snapshot in the chain
+      child = getSnapshotLocalData(
+          getSnapshotInfo(ozoneManager,
+              metadataManager.getSnapshotChainManager(),
+              parentId));
+    }
+    SnapshotInfo snapshotInfo =
+        getSnapshotInfo(ozoneManager, 
metadataManager.getSnapshotChainManager(), fromSnapshotId);
+    OmSnapshotLocalData fromSnapshot = getSnapshotLocalData(snapshotInfo);
+    // Get the version that the child was built from
+    OmSnapshotLocalData.VersionMeta childVersionMeta = 
child.getVersionSstFileInfos().get(child.getVersion());
+    int versionUsedWhenBuildingChild = 
childVersionMeta.getPreviousSnapshotVersion();
+    return 
fromSnapshot.getVersionSstFileInfos().get(versionUsedWhenBuildingChild);

Review Comment:
   This should not happen



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to