TrieLogPruner preload with 30 second timeout (#7365)

Also reduce pruning window from 30_000 to 5_000

---------

Signed-off-by: Simon Dudley <simon.dudley@consensys.net>
Co-authored-by: Sally MacFarlane <macfarla.github@gmail.com>

CHANGELOG.md

@@ -26,11 +26,13 @@
 - Add trie log pruner metrics [#7352](https://github.com/hyperledger/besu/pull/7352)
 - Force bonsai-limit-trie-logs-enabled=false when sync-mode=FULL instead of startup error [#7357](https://github.com/hyperledger/besu/pull/7357)
 - `--Xbonsai-parallel-tx-processing-enabled` option enables executing transactions in parallel during block processing for Bonsai nodes
+- Reduce default trie log pruning window size from 30,000 to 5,000 [#7365](https://github.com/hyperledger/besu/pull/7365)
 - Add option `--poa-discovery-retry-bootnodes` for PoA networks to always use bootnodes during peer refresh, not just on first start [#7314](https://github.com/hyperledger/besu/pull/7314) 
 
 ### Bug fixes
 - Fix `eth_call` deserialization to correctly ignore unknown fields in the transaction object. [#7323](https://github.com/hyperledger/besu/pull/7323)
 - Prevent Besu from starting up with sync-mode=FULL and bonsai-limit-trie-logs-enabled=true for private networks [#7357](https://github.com/hyperledger/besu/pull/7357)
+- Add 30 second timeout to trie log pruner preload [#7365](https://github.com/hyperledger/besu/pull/7365)
 - Avoid executing pruner preload during trie log subcommands [#7366](https://github.com/hyperledger/besu/pull/7366)
 
 ## 24.7.0

ethereum/core/src/main/java/org/hyperledger/besu/ethereum/trie/diffbased/common/trielog/TrieLogPruner.java

@@ -26,6 +26,12 @@ import org.hyperledger.besu.plugin.services.trielogs.TrieLogEvent;
 
 import java.util.Comparator;
 import java.util.Optional;
+import java.util.concurrent.Executors;
+import java.util.concurrent.ScheduledExecutorService;
+import java.util.concurrent.ScheduledFuture;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.TimeoutException;
+import java.util.concurrent.atomic.AtomicBoolean;
 import java.util.concurrent.atomic.AtomicLong;
 import java.util.function.Consumer;
 import java.util.stream.Stream;
@@ -40,6 +46,7 @@ import org.slf4j.LoggerFactory;
 public class TrieLogPruner implements TrieLogEvent.TrieLogObserver {
 
   private static final Logger LOG = LoggerFactory.getLogger(TrieLogPruner.class);
+  private static final int PRELOAD_TIMEOUT_IN_SECONDS = 30;
 
   private final int pruningLimit;
   private final int loadingLimit;
@@ -83,25 +90,60 @@ public class TrieLogPruner implements TrieLogEvent.TrieLogObserver {
             BesuMetricCategory.PRUNER, "trie_log_pruned_orphan", "trie log pruned orphan");
   }
 
-  public int initialize() {
-    return preloadQueue();
+  public void initialize() {
+    preloadQueueWithTimeout();
   }
 
-  private int preloadQueue() {
+  private void preloadQueueWithTimeout() {
+
     LOG.atInfo()
-        .setMessage("Loading first {} trie logs from database...")
+        .setMessage("Attempting to load first {} trie logs from database...")
         .addArgument(loadingLimit)
         .log();
+
+    try (final ScheduledExecutorService preloadExecutor = Executors.newScheduledThreadPool(1)) {
+
+      final AtomicBoolean timeoutOccurred = new AtomicBoolean(false);
+      final Runnable timeoutTask =
+          () -> {
+            timeoutOccurred.set(true);
+            LOG.atWarn()
+                .setMessage(
+                    "Timeout occurred while loading and processing {} trie logs from database")
+                .addArgument(loadingLimit)
+                .log();
+          };
+
+      final ScheduledFuture<?> timeoutFuture =
+          preloadExecutor.schedule(timeoutTask, PRELOAD_TIMEOUT_IN_SECONDS, TimeUnit.SECONDS);
+      LOG.atInfo()
+          .setMessage(
+              "Trie log pruning will timeout after {} seconds. If this is timing out, consider using `besu storage trie-log prune` subcommand, see https://besu.hyperledger.org/public-networks/how-to/bonsai-limit-trie-logs")
+          .addArgument(PRELOAD_TIMEOUT_IN_SECONDS)
+          .log();
+
+      preloadQueue(timeoutOccurred, timeoutFuture);
+    }
+  }
+
+  private void preloadQueue(
+      final AtomicBoolean timeoutOccurred, final ScheduledFuture<?> timeoutFuture) {
+
     try (final Stream<byte[]> trieLogKeys = rootWorldStateStorage.streamTrieLogKeys(loadingLimit)) {
-      final AtomicLong count = new AtomicLong();
+
+      final AtomicLong addToPruneQueueCount = new AtomicLong();
       final AtomicLong orphansPruned = new AtomicLong();
       trieLogKeys.forEach(
           blockHashAsBytes -> {
+            if (timeoutOccurred.get()) {
+              throw new RuntimeException(
+                  new TimeoutException("Timeout occurred while preloading trie log prune queue"));
+            }
             final Hash blockHash = Hash.wrap(Bytes32.wrap(blockHashAsBytes));
             final Optional<BlockHeader> header = blockchain.getBlockHeader(blockHash);
             if (header.isPresent()) {
               addToPruneQueue(header.get().getNumber(), blockHash);
-              count.getAndIncrement();
+              addToPruneQueueCount.getAndIncrement();
             } else {
               // prune orphaned blocks (sometimes created during block production)
               rootWorldStateStorage.pruneTrieLog(blockHash);
@@ -109,12 +151,21 @@ public class TrieLogPruner implements TrieLogEvent.TrieLogObserver {
               prunedOrphanCounter.inc();
             }
           });
+
+      timeoutFuture.cancel(true);
       LOG.atDebug().log("Pruned {} orphaned trie logs from database...", orphansPruned.intValue());
-      LOG.atInfo().log("Loaded {} trie logs from database", count);
-      return pruneFromQueue() + orphansPruned.intValue();
+      LOG.atInfo().log(
+          "Added {} trie logs to prune queue. Commencing pruning of eligible trie logs...",
+          addToPruneQueueCount.intValue());
+      int prunedCount = pruneFromQueue();
+      LOG.atInfo().log("Pruned {} trie logs.", prunedCount);
     } catch (Exception e) {
+      if (e.getCause() != null && e.getCause() instanceof TimeoutException) {
+        int prunedCount = pruneFromQueue();
+        LOG.atInfo().log("Operation timed out, but still pruned {} trie logs.", prunedCount);
+      } else {
         LOG.error("Error loading trie logs from database, nothing pruned", e);
-      return 0;
+      }
     }
   }
 

ethereum/core/src/main/java/org/hyperledger/besu/ethereum/worldstate/DataStorageConfiguration.java

@@ -25,7 +25,7 @@ public interface DataStorageConfiguration {
   long DEFAULT_BONSAI_MAX_LAYERS_TO_LOAD = 512;
   boolean DEFAULT_BONSAI_LIMIT_TRIE_LOGS_ENABLED = true;
   long MINIMUM_BONSAI_TRIE_LOG_RETENTION_LIMIT = DEFAULT_BONSAI_MAX_LAYERS_TO_LOAD;
-  int DEFAULT_BONSAI_TRIE_LOG_PRUNING_WINDOW_SIZE = 30_000;
+  int DEFAULT_BONSAI_TRIE_LOG_PRUNING_WINDOW_SIZE = 5_000;
   boolean DEFAULT_RECEIPT_COMPACTION_ENABLED = false;
 
   DataStorageConfiguration DEFAULT_CONFIG =