HDDS-8859. [Snapshot] Return failure message to client for a failed snapshot diff jobs

hadoop-hdds/common/src/main/resources/ozone-default.xml

@@ -3991,7 +3991,7 @@
 
   <property>
     <name>ozone.om.snapshot.diff.cleanup.service.run.internal</name>
-    <value>60m</value>
+    <value>1m</value>
     <tag>OZONE, OM</tag>
     <description>
       Interval at which snapshot diff clean up service will run.

hadoop-ozone/client/src/main/java/org/apache/hadoop/ozone/client/ObjectStore.java

@@ -45,6 +45,7 @@
 import org.apache.hadoop.ozone.om.helpers.TenantUserList;
 import org.apache.hadoop.ozone.security.OzoneTokenIdentifier;
 import org.apache.hadoop.ozone.security.acl.OzoneObj;
+import org.apache.hadoop.ozone.snapshot.CancelSnapshotDiffResponse;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffResponse;
 import org.apache.hadoop.security.UserGroupInformation;
 
@@ -667,22 +668,37 @@ private List<OzoneSnapshot> getNextListOfSnapshots(String prevSnapshot)
    * @param token to get the index to return diff report from.
    * @param pageSize maximum entries returned to the report.
    * @param forceFullDiff request to force full diff, skipping DAG optimization
-   * @param cancel request to cancel a running snapshot diff job.
    * @return the difference report between two snapshots
    * @throws IOException in case of any exception while generating snapshot diff
    */
-  @SuppressWarnings("parameternumber")
   public SnapshotDiffResponse snapshotDiff(String volumeName,
                                            String bucketName,
                                            String fromSnapshot,
                                            String toSnapshot,
                                            String token,
                                            int pageSize,
-                                           boolean forceFullDiff,
-                                           boolean cancel)
+                                           boolean forceFullDiff)
       throws IOException {
     return proxy.snapshotDiff(volumeName, bucketName, fromSnapshot, toSnapshot,
-        token, pageSize, forceFullDiff, cancel);
+        token, pageSize, forceFullDiff);
+  }
+
+  /**
+   * Cancel the snap diff jobs.
+   * @param volumeName Name of the volume to which the snapshot bucket belong
+   * @param bucketName Name of the bucket to which the snapshots belong
+   * @param fromSnapshot The name of the starting snapshot
+   * @param toSnapshot The name of the ending snapshot
+   * @return the success if cancel succeeds.
+   * @throws IOException in case of any exception while generating snapshot diff
+   */
+  public CancelSnapshotDiffResponse cancelSnapshotDiff(String volumeName,
+                                                       String bucketName,
+                                                       String fromSnapshot,
+                                                       String toSnapshot)
+      throws IOException {
+    return proxy.cancelSnapshotDiff(volumeName, bucketName, fromSnapshot,
+        toSnapshot);
   }
 
   /**

hadoop-ozone/client/src/main/java/org/apache/hadoop/ozone/client/protocol/ClientProtocol.java

@@ -64,6 +64,7 @@
 import org.apache.hadoop.ozone.protocol.proto.OzoneManagerProtocolProtos.OMRoleInfo;
 import org.apache.hadoop.ozone.security.OzoneTokenIdentifier;
 import org.apache.hadoop.ozone.security.acl.OzoneObj;
+import org.apache.hadoop.ozone.snapshot.CancelSnapshotDiffResponse;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffResponse;
 import org.apache.hadoop.security.KerberosInfo;
 import org.apache.hadoop.security.token.Token;
@@ -1058,7 +1059,6 @@ List<OzoneSnapshot> listSnapshot(
       String volumeName, String bucketName, String snapshotPrefix,
       String prevSnapshot, int maxListResult) throws IOException;
 
-
   /**
    * Get the differences between two snapshots.
    * @param volumeName Name of the volume to which the snapshotted bucket belong
@@ -1068,15 +1068,28 @@ List<OzoneSnapshot> listSnapshot(
    * @param token to get the index to return diff report from.
    * @param pageSize maximum entries returned to the report.
    * @param forceFullDiff request to force full diff, skipping DAG optimization
-   * @param cancel request to cancel a running snapshot diff job.
    * @return the difference report between two snapshots
    * @throws IOException in case of any exception while generating snapshot diff
    */
-  @SuppressWarnings("parameternumber")
   SnapshotDiffResponse snapshotDiff(String volumeName, String bucketName,
                                     String fromSnapshot, String toSnapshot,
                                     String token, int pageSize,
-                                    boolean forceFullDiff, boolean cancel)
+                                    boolean forceFullDiff)
+      throws IOException;
+
+  /**
+   * Cancel snapshot diff job.
+   * @param volumeName Name of the volume to which the snapshotted bucket belong
+   * @param bucketName Name of the bucket to which the snapshots belong
+   * @param fromSnapshot The name of the starting snapshot
+   * @param toSnapshot The name of the ending snapshot
+   * @return the success if cancel succeeds.
+   * @throws IOException in case of any exception while cancelling snap diff job
+   */
+  CancelSnapshotDiffResponse cancelSnapshotDiff(String volumeName,
+                                                String bucketName,
+                                                String fromSnapshot,
+                                                String toSnapshot)
       throws IOException;
 
   /**

hadoop-ozone/client/src/main/java/org/apache/hadoop/ozone/client/rpc/RpcClient.java

@@ -151,6 +151,7 @@
 import org.apache.hadoop.ozone.security.acl.IAccessAuthorizer.ACLType;
 import org.apache.hadoop.ozone.security.acl.OzoneAclConfig;
 import org.apache.hadoop.ozone.security.acl.OzoneObj;
+import org.apache.hadoop.ozone.snapshot.CancelSnapshotDiffResponse;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffResponse;
 import org.apache.hadoop.security.UserGroupInformation;
 import org.apache.hadoop.security.token.Token;
@@ -986,15 +987,32 @@ public SnapshotDiffResponse snapshotDiff(String volumeName,
                                            String toSnapshot,
                                            String token,
                                            int pageSize,
-                                           boolean forceFullDiff,
-                                           boolean cancel)
+                                           boolean forceFullDiff)
       throws IOException {
     Preconditions.checkArgument(StringUtils.isNotBlank(volumeName),
         "volume can't be null or empty.");
     Preconditions.checkArgument(StringUtils.isNotBlank(bucketName),
         "bucket can't be null or empty.");
     return ozoneManagerClient.snapshotDiff(volumeName, bucketName,
-        fromSnapshot, toSnapshot, token, pageSize, forceFullDiff, cancel);
+        fromSnapshot, toSnapshot, token, pageSize, forceFullDiff);
+  }
+
+  @Override
+  public CancelSnapshotDiffResponse cancelSnapshotDiff(String volumeName,
+                                                       String bucketName,
+                                                       String fromSnapshot,
+                                                       String toSnapshot)
+      throws IOException {
+    Preconditions.checkArgument(StringUtils.isNotBlank(volumeName),
+        "volume can't be null or empty.");
+    Preconditions.checkArgument(StringUtils.isNotBlank(bucketName),
+        "bucket can't be null or empty.");
+    Preconditions.checkArgument(StringUtils.isNotBlank(fromSnapshot),
+        "fromSnapshot can't be null or empty.");
+    Preconditions.checkArgument(StringUtils.isNotBlank(toSnapshot),
+        "toSnapshot can't be null or empty.");
+    return ozoneManagerClient.cancelSnapshotDiff(volumeName, bucketName,
+        fromSnapshot, toSnapshot);
   }
 
   @Override

hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/OmUtils.java

@@ -266,6 +266,7 @@ public static boolean isReadOnly(
       // operation SetRangerServiceVersion.
     case GetKeyInfo:
     case SnapshotDiff:
+    case CancelSnapshotDiff:
     case ListSnapshotDiffJobs:
     case TransferLeadership:
       return true;

hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/om/OMConfigKeys.java

@@ -527,7 +527,7 @@ private OMConfigKeys() {
       = "ozone.om.snapshot.diff.cleanup.service.run.internal";
   public static final long
       OZONE_OM_SNAPSHOT_DIFF_CLEANUP_SERVICE_RUN_INTERVAL_DEFAULT
-      = TimeUnit.HOURS.toMillis(1);
+      = TimeUnit.MINUTES.toMillis(1);
 
   public static final String OZONE_OM_SNAPSHOT_DIFF_CLEANUP_SERVICE_TIMEOUT
       = "ozone.om.snapshot.diff.cleanup.service.timeout";

hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/om/helpers/SnapshotDiffJob.java

@@ -22,6 +22,8 @@
 import com.fasterxml.jackson.databind.ObjectMapper;
 import java.io.IOException;
 import java.util.Objects;
+
+import org.apache.commons.lang3.StringUtils;
 import org.apache.hadoop.hdds.utils.db.Codec;
 import org.apache.hadoop.ozone.protocol.proto.OzoneManagerProtocolProtos.SnapshotDiffJobProto;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffResponse.JobStatus;
@@ -48,6 +50,10 @@ public static Codec<SnapshotDiffJob> getCodec() {
   private boolean forceFullDiff;
   private long totalDiffEntries;
 
+  // Reason tells why the job was FAILED. It should be set only if job status
+  // is FAILED.
+  private String reason;
+
   // Default constructor for Jackson Serializer.
   public SnapshotDiffJob() {
 
@@ -72,6 +78,7 @@ public SnapshotDiffJob(long creationTime,
     this.toSnapshot = toSnapshot;
     this.forceFullDiff = forceFullDiff;
     this.totalDiffEntries = totalDiffEntries;
+    this.reason = StringUtils.EMPTY;
   }
 
   public String getJobId() {
@@ -146,17 +153,30 @@ public void setTotalDiffEntries(long totalDiffEntries) {
     this.totalDiffEntries = totalDiffEntries;
   }
 
+  public String getReason() {
+    return reason;
+  }
+
+  public void setReason(String reason) {
+    this.reason = reason;
+  }
+
   @Override
   public String toString() {
-    return "creationTime : " + creationTime +
-        ", jobId: " + jobId +
-        ", status: " + status +
-        ", volume: " + volume +
-        ", bucket: " + bucket +
-        ", fromSnapshot: " + fromSnapshot +
-        ", toSnapshot: " + toSnapshot +
-        ", forceFullDiff: " + forceFullDiff +
-        ", totalDiffEntries: " + totalDiffEntries;
+    StringBuilder sb = new StringBuilder("creationTime : ").append(creationTime)
+        .append(", jobId: ").append(jobId)
+        .append(", status: ").append(status)
+        .append(", volume: ").append(volume)
+        .append(", bucket: ").append(bucket)
+        .append(", fromSnapshot: ").append(fromSnapshot)
+        .append(", toSnapshot: ").append(toSnapshot)
+        .append(", forceFullDiff: ").append(forceFullDiff)
+        .append(", totalDiffEntries: ").append(totalDiffEntries);
+
+    if (StringUtils.isNotEmpty(reason)) {
+      sb.append(", reason: ").append(reason);
+    }
+    return sb.toString();
   }
 
   @Override
@@ -175,15 +195,16 @@ public boolean equals(Object other) {
           Objects.equals(this.fromSnapshot, otherJob.fromSnapshot) &&
           Objects.equals(this.toSnapshot, otherJob.toSnapshot) &&
           Objects.equals(this.forceFullDiff, otherJob.forceFullDiff) &&
-          Objects.equals(this.totalDiffEntries, otherJob.totalDiffEntries);
+          Objects.equals(this.totalDiffEntries, otherJob.totalDiffEntries) &&
+          Objects.equals(this.reason, otherJob.reason);
     }
     return false;
   }
 
   @Override
   public int hashCode() {
     return Objects.hash(creationTime, jobId, status, volume, bucket,
-        fromSnapshot, toSnapshot, forceFullDiff, totalDiffEntries);
+        fromSnapshot, toSnapshot, forceFullDiff, totalDiffEntries, reason);
   }
 
   public SnapshotDiffJobProto toProtoBuf() {

hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/om/protocol/OzoneManagerProtocol.java

@@ -66,6 +66,7 @@
 import org.apache.hadoop.ozone.protocol.proto.OzoneManagerProtocolProtos.EchoRPCResponse;
 import org.apache.hadoop.ozone.security.OzoneDelegationTokenSelector;
 import org.apache.hadoop.ozone.security.acl.OzoneObj;
+import org.apache.hadoop.ozone.snapshot.CancelSnapshotDiffResponse;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffResponse;
 import org.apache.hadoop.ozone.upgrade.UpgradeFinalizer.StatusAndMessages;
 import org.apache.hadoop.security.KerberosInfo;
@@ -715,19 +716,34 @@ default List<SnapshotInfo> listSnapshot(
    * @param token to get the index to return diff report from.
    * @param pageSize maximum entries returned to the report.
    * @param forceFullDiff request to force full diff, skipping DAG optimization
-   * @param cancel request to cancel a running snapshot diff job.
    * @return the difference report between two snapshots
    * @throws IOException in case of any exception while generating snapshot diff
    */
-  @SuppressWarnings("parameternumber")
   default SnapshotDiffResponse snapshotDiff(String volumeName,
                                             String bucketName,
                                             String fromSnapshot,
                                             String toSnapshot,
                                             String token,
                                             int pageSize,
-                                            boolean forceFullDiff,
-                                            boolean cancel)
+                                            boolean forceFullDiff)
+      throws IOException {
+    throw new UnsupportedOperationException("OzoneManager does not require " +
+        "this to be implemented");
+  }
+
+  /**
+   * Cancel snapshot diff job.
+   * @param volumeName Name of the volume to which the snapshotted bucket belong
+   * @param bucketName Name of the bucket to which the snapshots belong
+   * @param fromSnapshot The name of the starting snapshot
+   * @param toSnapshot The name of the ending snapshot
+   * @return the success if cancel succeeds.
+   * @throws IOException in case of any exception while cancelling snap diff job
+   */
+  default CancelSnapshotDiffResponse cancelSnapshotDiff(String volumeName,
+                                                        String bucketName,
+                                                        String fromSnapshot,
+                                                        String toSnapshot)
       throws IOException {
     throw new UnsupportedOperationException("OzoneManager does not require " +
         "this to be implemented");

hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/om/protocolPB/OzoneManagerProtocolClientSideTranslatorPB.java

@@ -196,10 +196,10 @@
 import org.apache.hadoop.ozone.security.proto.SecurityProtos.CancelDelegationTokenRequestProto;
 import org.apache.hadoop.ozone.security.proto.SecurityProtos.GetDelegationTokenRequestProto;
 import org.apache.hadoop.ozone.security.proto.SecurityProtos.RenewDelegationTokenRequestProto;
+import org.apache.hadoop.ozone.snapshot.CancelSnapshotDiffResponse;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffReportOzone;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffResponse;
 import org.apache.hadoop.ozone.snapshot.SnapshotDiffResponse.JobStatus;
-import org.apache.hadoop.ozone.snapshot.SnapshotDiffResponse.JobCancelResult;
 import org.apache.hadoop.ozone.upgrade.UpgradeFinalizer;
 import org.apache.hadoop.ozone.upgrade.UpgradeFinalizer.StatusAndMessages;
 import org.apache.hadoop.security.token.Token;
@@ -1217,8 +1217,7 @@ public SnapshotDiffResponse snapshotDiff(String volumeName,
                                            String toSnapshot,
                                            String token,
                                            int pageSize,
-                                           boolean forceFullDiff,
-                                           boolean cancel)
+                                           boolean forceFullDiff)
       throws IOException {
     final OzoneManagerProtocolProtos.SnapshotDiffRequest.Builder
         requestBuilder =
@@ -1228,8 +1227,7 @@ public SnapshotDiffResponse snapshotDiff(String volumeName,
             .setFromSnapshot(fromSnapshot)
             .setToSnapshot(toSnapshot)
             .setPageSize(pageSize)
-            .setForceFullDiff(forceFullDiff)
-            .setCancel(cancel);
+            .setForceFullDiff(forceFullDiff);
 
     if (!StringUtils.isBlank(token)) {
       requestBuilder.setToken(token);
@@ -1247,7 +1245,36 @@ public SnapshotDiffResponse snapshotDiff(String volumeName,
         diffResponse.getSnapshotDiffReport()),
         JobStatus.fromProtobuf(diffResponse.getJobStatus()),
         diffResponse.getWaitTimeInMs(),
-        JobCancelResult.fromProtobuf(diffResponse.getJobCancelResult()));
+        diffResponse.getReason());
+  }
+
+  /**
+   * {@inheritDoc}
+   */
+  @Override
+  public CancelSnapshotDiffResponse cancelSnapshotDiff(String volumeName,
+                                                       String bucketName,
+                                                       String fromSnapshot,
+                                                       String toSnapshot)
+      throws IOException {
+    final OzoneManagerProtocolProtos.CancelSnapshotDiffRequest.Builder
+        requestBuilder =
+        OzoneManagerProtocolProtos.CancelSnapshotDiffRequest.newBuilder()
+            .setVolumeName(volumeName)
+            .setBucketName(bucketName)
+            .setFromSnapshot(fromSnapshot)
+            .setToSnapshot(toSnapshot);
+
+    final OMRequest omRequest = createOMRequest(Type.CancelSnapshotDiff)
+        .setCancelSnapshotDiffRequest(requestBuilder)
+        .build();
+
+    final OMResponse omResponse = submitRequest(omRequest);
+    handleError(omResponse);
+    OzoneManagerProtocolProtos.CancelSnapshotDiffResponse diffResponse =
+        omResponse.getCancelSnapshotDiffResponse();
+
+    return new CancelSnapshotDiffResponse(diffResponse.getReason());
   }
 
   /**