HDFS-11289. [SPS]: Make SPS movement monitor timeouts configurable. Contributed by Uma Maheswara Rao G

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSConfigKeys.java

@@ -613,10 +613,19 @@ public class DFSConfigKeys extends CommonConfigurationKeys {
   public static final String  DFS_MOVER_MAX_NO_MOVE_INTERVAL_KEY = "dfs.mover.max-no-move-interval";
   public static final int    DFS_MOVER_MAX_NO_MOVE_INTERVAL_DEFAULT = 60*1000; // One minute
 
+  // SPS related configurations
   public static final String  DFS_STORAGE_POLICY_SATISFIER_ACTIVATE_KEY =
       "dfs.storage.policy.satisfier.activate";
   public static final boolean DFS_STORAGE_POLICY_SATISFIER_ACTIVATE_DEFAULT =
       true;
+  public static final String DFS_STORAGE_POLICY_SATISFIER_RECHECK_TIMEOUT_MILLIS_KEY =
+      "dfs.storage.policy.satisfier.recheck.timeout.millis";
+  public static final int DFS_STORAGE_POLICY_SATISFIER_RECHECK_TIMEOUT_MILLIS_DEFAULT =
+      5 * 60 * 1000;
+  public static final String DFS_STORAGE_POLICY_SATISFIER_SELF_RETRY_TIMEOUT_MILLIS_KEY =
+      "dfs.storage.policy.satisfier.self.retry.timeout.millis";
+  public static final int DFS_STORAGE_POLICY_SATISFIER_SELF_RETRY_TIMEOUT_MILLIS_DEFAULT =
+      30 * 60 * 1000;
 
   public static final String  DFS_DATANODE_ADDRESS_KEY = "dfs.datanode.address";
   public static final int     DFS_DATANODE_DEFAULT_PORT = 9866;

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java

@@ -479,8 +479,8 @@ public class BlockManager implements BlockStatsMXBean {
             DFSConfigKeys.DFS_STORAGE_POLICY_SATISFIER_ACTIVATE_KEY,
             DFSConfigKeys.DFS_STORAGE_POLICY_SATISFIER_ACTIVATE_DEFAULT);
     if (storagePolicyEnabled && spsEnabled) {
-      sps = new StoragePolicySatisfier(namesystem,
+      sps = new StoragePolicySatisfier(namesystem, storageMovementNeeded, this,
-          storageMovementNeeded, this);
+          conf);
     } else {
       sps = null;
       LOG.warn(

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/BlockStorageMovementAttemptedItems.java

@@ -64,14 +64,14 @@ public class BlockStorageMovementAttemptedItems {
   // It might take anywhere between 5 to 10 minutes before
   // a request is timed out.
   //
-  private long checkTimeout = 5 * 60 * 1000; // minimum value
+  private long minCheckTimeout = 5 * 60 * 1000; // minimum value
   private BlockStorageMovementNeeded blockStorageMovementNeeded;
 
-  public BlockStorageMovementAttemptedItems(long timeoutPeriod,
+  public BlockStorageMovementAttemptedItems(long recheckTimeout,
       long selfRetryTimeout,
       BlockStorageMovementNeeded unsatisfiedStorageMovementFiles) {
-    if (timeoutPeriod > 0) {
+    if (recheckTimeout > 0) {
-      this.checkTimeout = Math.min(checkTimeout, timeoutPeriod);
+      this.minCheckTimeout = Math.min(minCheckTimeout, recheckTimeout);
     }
 
     this.selfRetryTimeout = selfRetryTimeout;
@@ -196,7 +196,7 @@ public class BlockStorageMovementAttemptedItems {
         try {
           blockStorageMovementResultCheck();
           blocksStorageMovementUnReportedItemsCheck();
-          Thread.sleep(checkTimeout);
+          Thread.sleep(minCheckTimeout);
         } catch (InterruptedException ie) {
           LOG.info("BlocksStorageMovementAttemptResultMonitor thread "
               + "is interrupted.", ie);

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/StoragePolicySatisfier.java

@@ -27,7 +27,9 @@ import java.util.LinkedList;
 import java.util.List;
 
 import org.apache.hadoop.classification.InterfaceAudience;
+import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.StorageType;
+import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DFSUtil;
 import org.apache.hadoop.hdfs.protocol.Block;
 import org.apache.hadoop.hdfs.protocol.BlockStoragePolicy;
@@ -79,15 +81,18 @@ public class StoragePolicySatisfier implements Runnable {
 
   public StoragePolicySatisfier(final Namesystem namesystem,
       final BlockStorageMovementNeeded storageMovementNeeded,
-      final BlockManager blkManager) {
+      final BlockManager blkManager, Configuration conf) {
     this.namesystem = namesystem;
     this.storageMovementNeeded = storageMovementNeeded;
     this.blockManager = blkManager;
-    // TODO: below selfRetryTimeout and checkTimeout can be configurable later
-    // Now, the default values of selfRetryTimeout and checkTimeout are 30mins
-    // and 5mins respectively
     this.storageMovementsMonitor = new BlockStorageMovementAttemptedItems(
-        5 * 60 * 1000, 30 * 60 * 1000, storageMovementNeeded);
+        conf.getLong(
+            DFSConfigKeys.DFS_STORAGE_POLICY_SATISFIER_RECHECK_TIMEOUT_MILLIS_KEY,
+            DFSConfigKeys.DFS_STORAGE_POLICY_SATISFIER_RECHECK_TIMEOUT_MILLIS_DEFAULT),
+        conf.getLong(
+            DFSConfigKeys.DFS_STORAGE_POLICY_SATISFIER_SELF_RETRY_TIMEOUT_MILLIS_KEY,
+            DFSConfigKeys.DFS_STORAGE_POLICY_SATISFIER_SELF_RETRY_TIMEOUT_MILLIS_DEFAULT),
+        storageMovementNeeded);
   }
 
   /**

hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml

@@ -4509,6 +4509,29 @@
   </description>
 </property>
 
+<property>
+  <name>dfs.storage.policy.satisfier.recheck.timeout.millis</name>
+  <value>300000</value>
+  <description>
+    Blocks storage movements monitor re-check interval in milliseconds.
+    This check will verify whether any blocks storage movement results arrived from DN
+    and also verify if any of file blocks movements not at all reported to DN
+    since dfs.storage.policy.satisfier.self.retry.timeout.
+    The default value is 5 * 60 * 1000 (5 mins)
+  </description>
+</property>
+
+<property>
+  <name>dfs.storage.policy.satisfier.self.retry.timeout.millis</name>
+  <value>1800000</value>
+  <description>
+    If any of file related block movements not at all reported by coordinator datanode,
+    then after this timeout(in milliseconds), the item will be added back to movement needed list
+    at namenode which will be retried for block movements.
+    The default value is 30 * 60 * 1000 (30 mins)
+  </description>
+</property>
+
 <property>
   <name>dfs.pipeline.ecn</name>
   <value>false</value>