NIFI-15570: Keep track of Content Claims where the last Claim in a Resource Claim can be truncated if it is large. Whenever FlowFile Repository is checkpointed, truncate any large Resource Claims when possible and necessary to avoid having a situtation where a small FlowFile in a given Resource Claim prevents a large Content Claim from being cleaned up.

Author: markap14

nifi-framework-api/src/main/java/org/apache/nifi/controller/repository/claim/ContentClaim.java

@@ -44,4 +44,12 @@ public interface ContentClaim extends Comparable<ContentClaim> {
      * @return the length of this ContentClaim
      */
     long getLength();
+
+    /**
+     * Indicates whether or not this ContentClaim is a candidate for truncation.
+     * @return true if this ContentClaim is a candidate for truncation, false otherwise
+     */
+    default boolean isTruncationCandidate() {
+        return false;
+    }
 }

nifi-framework-api/src/main/java/org/apache/nifi/controller/repository/claim/ResourceClaimManager.java

@@ -113,6 +113,17 @@ public interface ResourceClaimManager {
      */
     void markDestructable(ResourceClaim claim);
 
+    /**
+     * Indicates that the Resource Claim associated with the given Content Claim can now be
+     * truncated to the start of the ContentClaim. This should only ever be called after it is
+     * guaranteed that the FlowFile Repository has been synchronized with its underlying
+     * storage component for the same reason as described in the {@link #markDestructable(ResourceClaim)}
+     * method.
+     *
+     * @param claim the ContentClaim that should be used for truncation
+     */
+    void markTruncatable(ContentClaim claim);
+
     /**
      * Drains up to {@code maxElements} Content Claims from the internal queue
      * of destructable content claims to the given {@code destination} so that
@@ -139,6 +150,16 @@ public interface ResourceClaimManager {
      */
     void drainDestructableClaims(Collection<ResourceClaim> destination, int maxElements, long timeout, TimeUnit unit);
 
+    /**
+     * Drains up to {@code maxElements} Content Claims from the internal queue
+     * of truncatable content claims to the given {@code destination} so that
+     * they can be truncated.
+     *
+     * @param destination to drain to
+     * @param maxElements max items to drain
+     */
+    void drainTruncatableClaims(Collection<ContentClaim> destination, int maxElements);
+
     /**
      * Clears the manager's memory of any and all ResourceClaims that it knows
      * about

nifi-framework-bundle/nifi-framework/nifi-framework-core/src/main/java/org/apache/nifi/controller/repository/FileSystemRepository.java

@@ -21,7 +21,9 @@
 import org.apache.nifi.controller.repository.claim.ContentClaim;
 import org.apache.nifi.controller.repository.claim.ResourceClaim;
 import org.apache.nifi.controller.repository.claim.ResourceClaimManager;
+import org.apache.nifi.controller.repository.claim.StandardContentClaim;
 import org.apache.nifi.flowfile.attributes.CoreAttributes;
+import org.apache.nifi.processor.DataUnit;
 import org.apache.nifi.repository.schema.FieldCache;
 import org.apache.nifi.util.FormatUtils;
 import org.apache.nifi.util.NiFiProperties;
@@ -98,6 +100,7 @@ public class WriteAheadFlowFileRepository implements FlowFileRepository, SyncLis
     private final List<File> flowFileRepositoryPaths = new ArrayList<>();
     private final ScheduledExecutorService checkpointExecutor;
     private final int maxCharactersToCache;
+    private final long truncationThreshold;
 
     private volatile Collection<SerializedRepositoryRecord> recoveredRecords = null;
     private final Set<ResourceClaim> orphanedResourceClaims = Collections.synchronizedSet(new HashSet<>());
@@ -132,6 +135,7 @@ public class WriteAheadFlowFileRepository implements FlowFileRepository, SyncLis
     // before the data is destroyed, it's okay because the data will be unknown to the Content Repository, so it will be destroyed
     // on restart.
     private final ConcurrentMap<Integer, BlockingQueue<ResourceClaim>> claimsAwaitingDestruction = new ConcurrentHashMap<>();
+    private final ConcurrentMap<Integer, BlockingQueue<ContentClaim>> claimsAwaitingTruncation = new ConcurrentHashMap<>();
 
     /**
      * default no args constructor for service loading only.
@@ -143,6 +147,7 @@ public WriteAheadFlowFileRepository() {
         nifiProperties = null;
         retainOrphanedFlowFiles = true;
         maxCharactersToCache = 0;
+        truncationThreshold = Long.MAX_VALUE;
     }
 
     public WriteAheadFlowFileRepository(final NiFiProperties nifiProperties) {
@@ -153,6 +158,8 @@ public WriteAheadFlowFileRepository(final NiFiProperties nifiProperties) {
         retainOrphanedFlowFiles = orphanedFlowFileProperty == null || Boolean.parseBoolean(orphanedFlowFileProperty);
 
         this.maxCharactersToCache = nifiProperties.getIntegerProperty(FLOWFILE_REPO_CACHE_SIZE, DEFAULT_CACHE_SIZE);
+        final long maxAppendableClaimLength = DataUnit.parseDataSize(nifiProperties.getMaxAppendableClaimSize(), DataUnit.B).longValue();
+        truncationThreshold = Math.min(1_000_000, maxAppendableClaimLength);
 
         final String directoryName = nifiProperties.getProperty(FLOWFILE_REPOSITORY_DIRECTORY_PREFIX);
         flowFileRepositoryPaths.add(new File(directoryName));
@@ -446,33 +453,48 @@ protected void updateContentClaims(Collection<RepositoryRecord> repositoryRecord
         // The below code is not entirely thread-safe, but we are OK with that because the results aren't really harmful.
         // Specifically, if two different threads call updateRepository with DELETE records for the same Content Claim,
         // it's quite possible for claimant count to be 0 below, which results in two different threads adding the Content
-        // Claim to the 'claimsAwaitDestruction' map. As a result, we can call #markDestructable with the same ContentClaim
+        // Claim to the 'claimsAwaitingDestruction' map. As a result, we can call #markDestructable with the same ContentClaim
         // multiple times, and the #markDestructable method is not necessarily idempotent.
         // However, the result of this is that the FileSystem Repository may end up trying to remove the content multiple times.
         // This does not, however, cause problems, as ContentRepository should handle this
         // This does indicate that some refactoring should probably be performed, though, as this is not a very clean interface.
-        final Set<ResourceClaim> claimsToAdd = new HashSet<>();
+        final Set<ResourceClaim> destructableClaims = new HashSet<>();
+        final Set<ContentClaim> truncatableClaims = new HashSet<>();
 
         final Set<String> swapLocationsAdded = new HashSet<>();
         final Set<String> swapLocationsRemoved = new HashSet<>();
 
         for (final RepositoryRecord record : repositoryRecords) {
             updateClaimCounts(record);
 
+            final ContentClaim contentClaim = record.getCurrentClaim();
+            final boolean truncationCandidate = contentClaim != null && contentClaim.isTruncationCandidate();
+            final boolean claimChanged = !Objects.equals(record.getOriginalClaim(), contentClaim);
             if (record.getType() == RepositoryRecordType.DELETE) {
-                // For any DELETE record that we have, if claim is destructible, mark it so
-                if (record.getCurrentClaim() != null && isDestructable(record.getCurrentClaim())) {
-                    claimsToAdd.add(record.getCurrentClaim().getResourceClaim());
+                // For any DELETE record that we have, if claim is destructible or truncatable, mark it so
+                if (isDestructable(contentClaim)) {
+                    destructableClaims.add(contentClaim.getResourceClaim());
+                } else if (truncationCandidate) {
+                    truncatableClaims.add(contentClaim);
                 }
 
-                // If the original claim is different than the current claim and the original claim is destructible, mark it so
-                if (record.getOriginalClaim() != null && !record.getOriginalClaim().equals(record.getCurrentClaim()) && isDestructable(record.getOriginalClaim())) {
-                    claimsToAdd.add(record.getOriginalClaim().getResourceClaim());
+                // If the original claim is different than the current claim and the original claim is destructible
+                // or truncatable, mark it so
+                if (claimChanged) {
+                    if (isDestructable(record.getOriginalClaim())) {
+                        destructableClaims.add(record.getOriginalClaim().getResourceClaim());
+                    } else if (record.getOriginalClaim() != null && record.getOriginalClaim().isTruncationCandidate()) {
+                        truncatableClaims.add(record.getOriginalClaim());
+                    }
                 }
             } else if (record.getType() == RepositoryRecordType.UPDATE) {
                 // if we have an update, and the original is no longer needed, mark original as destructible
-                if (record.getOriginalClaim() != null && record.getCurrentClaim() != record.getOriginalClaim() && isDestructable(record.getOriginalClaim())) {
-                    claimsToAdd.add(record.getOriginalClaim().getResourceClaim());
+                if (claimChanged) {
+                    if (isDestructable(record.getOriginalClaim())) {
+                        destructableClaims.add(record.getOriginalClaim().getResourceClaim());
+                    } else if (record.getOriginalClaim() != null && record.getOriginalClaim().isTruncationCandidate()) {
+                        truncatableClaims.add(record.getOriginalClaim());
+                    }
                 }
             } else if (record.getType() == RepositoryRecordType.SWAP_OUT) {
                 final String swapLocation = record.getSwapLocation();
@@ -485,13 +507,16 @@ protected void updateContentClaims(Collection<RepositoryRecord> repositoryRecord
             }
         }
 
-        // Once the content claim counts have been updated for all records, collect any transient claims that are eligible for destruction
+        // Once the content claim counts have been updated for all records, collect any transient
+        // claims that are eligible for destruction or truncation
         for (final RepositoryRecord record : repositoryRecords) {
             final List<ContentClaim> transientClaims = record.getTransientClaims();
             if (transientClaims != null) {
                 for (final ContentClaim transientClaim : transientClaims) {
                     if (isDestructable(transientClaim)) {
-                        claimsToAdd.add(transientClaim.getResourceClaim());
+                        destructableClaims.add(transientClaim.getResourceClaim());
+                    } else if (transientClaim.isTruncationCandidate()) {
+                        truncatableClaims.add(transientClaim);
                     }
                 }
             }
@@ -505,19 +530,15 @@ protected void updateContentClaims(Collection<RepositoryRecord> repositoryRecord
             }
         }
 
-        if (!claimsToAdd.isEmpty()) {
-            // Get / Register a Set<ContentClaim> for the given Partition Index
-            final Integer partitionKey = Integer.valueOf(partitionIndex);
-            BlockingQueue<ResourceClaim> claimQueue = claimsAwaitingDestruction.get(partitionKey);
-            if (claimQueue == null) {
-                claimQueue = new LinkedBlockingQueue<>();
-                final BlockingQueue<ResourceClaim> existingClaimQueue = claimsAwaitingDestruction.putIfAbsent(partitionKey, claimQueue);
-                if (existingClaimQueue != null) {
-                    claimQueue = existingClaimQueue;
-                }
-            }
+        if (!destructableClaims.isEmpty()) {
+            // Get / Register a Set<ResourceClaim> for the given Partition Index
+            final BlockingQueue<ResourceClaim> claimQueue = claimsAwaitingDestruction.computeIfAbsent(partitionIndex, key -> new LinkedBlockingQueue<>());
+            claimQueue.addAll(destructableClaims);
+        }
 
-            claimQueue.addAll(claimsToAdd);
+        if (!truncatableClaims.isEmpty()) {
+            final BlockingQueue<ContentClaim> claimQueue = claimsAwaitingTruncation.computeIfAbsent(partitionIndex, key -> new LinkedBlockingQueue<>());
+            claimQueue.addAll(truncatableClaims);
         }
     }
 
@@ -568,16 +589,24 @@ private static String getLocationSuffix(final String swapLocation) {
 
     @Override
     public void onSync(final int partitionIndex) {
-        final BlockingQueue<ResourceClaim> claimQueue = claimsAwaitingDestruction.get(partitionIndex);
-        if (claimQueue == null) {
-            return;
+        final BlockingQueue<ResourceClaim> destructionClaimQueue = claimsAwaitingDestruction.get(partitionIndex);
+        if (destructionClaimQueue != null) {
+            final Set<ResourceClaim> claimsToDestroy = new HashSet<>();
+            destructionClaimQueue.drainTo(claimsToDestroy);
+
+            for (final ResourceClaim claim : claimsToDestroy) {
+                markDestructable(claim);
+            }
         }
 
-        final Set<ResourceClaim> claimsToDestroy = new HashSet<>();
-        claimQueue.drainTo(claimsToDestroy);
+        final BlockingQueue<ContentClaim> truncationClaimQueue = claimsAwaitingTruncation.get(partitionIndex);
+        if (truncationClaimQueue != null) {
+            final Set<ContentClaim> claimsToTruncate = new HashSet<>();
+            truncationClaimQueue.drainTo(claimsToTruncate);
 
-        for (final ResourceClaim claim : claimsToDestroy) {
-            markDestructable(claim);
+            for (final ContentClaim claim : claimsToTruncate) {
+                claimManager.markTruncatable(claim);
+            }
         }
     }
 
@@ -591,6 +620,15 @@ public void onGlobalSync() {
                 markDestructable(claim);
             }
         }
+
+        for (final BlockingQueue<ContentClaim> claimQueue : claimsAwaitingTruncation.values()) {
+            final Set<ContentClaim> claimsToTruncate = new HashSet<>();
+            claimQueue.drainTo(claimsToTruncate);
+
+            for (final ContentClaim claim : claimsToTruncate) {
+                claimManager.markTruncatable(claim);
+            }
+        }
     }
 
     /**
@@ -725,6 +763,10 @@ public long loadFlowFiles(final QueueProvider queueProvider) throws IOException
             queueMap.put(queue.getIdentifier(), queue);
         }
 
+        final Set<StandardContentClaim> truncationEligibleClaims = new HashSet<>();
+        final Set<ContentClaim> forbiddenTruncationClaims = new HashSet<>();
+        final Map<ResourceClaim, ContentClaim> latestContentClaimByResourceClaim = new HashMap<>();
+
         final List<SerializedRepositoryRecord> dropRecords = new ArrayList<>();
         int numFlowFilesMissingQueue = 0;
         long maxId = 0;
@@ -750,6 +792,15 @@ public long loadFlowFiles(final QueueProvider queueProvider) throws IOException
             }
 
             final ContentClaim claim = record.getContentClaim();
+
+            // Track the latest Content Claim for each Resource Claim so that we can determine which claims are eligible for truncation.
+            if (claim != null) {
+                final ContentClaim latestContentClaim = latestContentClaimByResourceClaim.get(claim.getResourceClaim());
+                if (latestContentClaim == null || claim.getOffset() > latestContentClaim.getOffset()) {
+                    latestContentClaimByResourceClaim.put(claim.getResourceClaim(), claim);
+                }
+            }
+
             final FlowFileQueue flowFileQueue = queueMap.get(queueId);
             final boolean orphaned = flowFileQueue == null;
             if (orphaned) {
@@ -779,6 +830,18 @@ public long loadFlowFiles(final QueueProvider queueProvider) throws IOException
 
                 continue;
             } else if (claim != null) {
+                // If the claim exceeds the max appendable claim length on its own and doesn't start the Resource Claim,
+                // we will consider it to be eligible for truncation. However, if there are multiple FlowFiles sharing the
+                // same claim, we cannot truncate it because doing so would affect the other FlowFiles.
+                if (claim.getOffset() > 0 && claim.getLength() > truncationThreshold && claim instanceof final StandardContentClaim scc) {
+                    if (forbiddenTruncationClaims.contains(claim) || truncationEligibleClaims.contains(scc)) {
+                        truncationEligibleClaims.remove(scc);
+                        forbiddenTruncationClaims.add(scc);
+                    } else {
+                        truncationEligibleClaims.add(scc);
+                    }
+                }
+
                 claimManager.incrementClaimantCount(claim.getResourceClaim());
             }
 
@@ -788,6 +851,16 @@ public long loadFlowFiles(final QueueProvider queueProvider) throws IOException
         // If recoveredRecords has been populated it need to be nulled out now because it is no longer useful and can be garbage collected.
         recoveredRecords = null;
 
+        // If any Content Claim was determined to be truncatable, mark it as such now.
+        for (final StandardContentClaim eligible : truncationEligibleClaims) {
+            final ContentClaim latestForResource = latestContentClaimByResourceClaim.get(eligible.getResourceClaim());
+            if (!Objects.equals(eligible, latestForResource)) {
+                continue;
+            }
+
+            eligible.setTruncationCandidate(true);
+        }
+
         // Set the AtomicLong to 1 more than the max ID so that calls to #getNextFlowFileSequence() will
         // return the appropriate number.
         flowFileSequenceGenerator.set(maxId + 1);
@@ -854,7 +927,7 @@ public long getNextFlowFileSequence() {
     }
 
     @Override
-    public long getMaxFlowFileIdentifier() throws IOException {
+    public long getMaxFlowFileIdentifier() {
         // flowFileSequenceGenerator is 1 more than the MAX so that we can call #getAndIncrement on the AtomicLong
         return flowFileSequenceGenerator.get() - 1;
     }

nifi-framework-bundle/nifi-framework/nifi-framework-core/src/main/java/org/apache/nifi/controller/repository/WriteAheadFlowFileRepository.java

@@ -21,6 +21,7 @@
 import org.apache.nifi.controller.repository.FlowFileRepository;
 import org.apache.nifi.controller.repository.SwapContents;
 import org.apache.nifi.controller.repository.SwapManagerInitializationContext;
+import org.apache.nifi.controller.repository.claim.ContentClaim;
 import org.apache.nifi.controller.repository.claim.ResourceClaim;
 import org.apache.nifi.controller.repository.claim.ResourceClaimManager;
 import org.apache.nifi.events.EventReporter;
@@ -223,6 +224,10 @@ public int incrementClaimantCount(ResourceClaim claim, boolean newClaim) {
         public void markDestructable(ResourceClaim claim) {
         }
 
+        @Override
+        public void markTruncatable(final ContentClaim claim) {
+        }
+
         @Override
         public void drainDestructableClaims(Collection<ResourceClaim> destination, int maxElements) {
         }
@@ -231,6 +236,10 @@ public void drainDestructableClaims(Collection<ResourceClaim> destination, int m
         public void drainDestructableClaims(Collection<ResourceClaim> destination, int maxElements, long timeout, TimeUnit unit) {
         }
 
+        @Override
+        public void drainTruncatableClaims(final Collection<ContentClaim> destination, final int maxElements) {
+        }
+
         @Override
         public void purge() {
         }