(accumulo) branch 2.1 updated: Defers server invalidation in tablet location cache (#5308)

[kturner]

This is an automated email from the ASF dual-hosted git repository.

kturner pushed a commit to branch 2.1
in repository https://gitbox.apache.org/repos/asf/accumulo.git


The following commit(s) were added to refs/heads/2.1 by this push:
     new 7d51553dfb Defers server invalidation in tablet location cache (#5308)
7d51553dfb is described below

commit 7d51553dfb20f3dc67015d45e5e3ad1b4ee770f9
Author: Keith Turner <ktur...@apache.org>
AuthorDate: Fri Feb 7 11:19:43 2025 -0500

    Defers server invalidation in tablet location cache (#5308)
    
    When many accumulo client threads all decided to invalidate some servers
    at the same time in the tablet location cache each threads would scan
    though all metadata looking for those servers.
    
    This change adds servers to a set for deferred processing when
    invalidated.  This has two advantages.  First for multiple threads
    adding servers at around the same time it can lower the number of scans
    done over all the cached tablet locations.  Second the amount of work
    done to invalidate servers is lowered.  The code used to find cache
    entries for a server, add the extent for that cache entry to badExtents,
    and then later look for everything in badExtents and remove it.  This
    change removes all of that work and directly removes entries from the
    cache and avoids adding anything to badExtents.
---
 .../core/clientImpl/TabletLocatorImpl.java         | 37 +++++++++++++++-------
 1 file changed, 25 insertions(+), 12 deletions(-)

diff --git 
a/core/src/main/java/org/apache/accumulo/core/clientImpl/TabletLocatorImpl.java 
b/core/src/main/java/org/apache/accumulo/core/clientImpl/TabletLocatorImpl.java
index d1e935577b..d46f7c9ef4 100644
--- 
a/core/src/main/java/org/apache/accumulo/core/clientImpl/TabletLocatorImpl.java
+++ 
b/core/src/main/java/org/apache/accumulo/core/clientImpl/TabletLocatorImpl.java
@@ -86,6 +86,7 @@ public class TabletLocatorImpl extends TabletLocator {
   protected Text lastTabletRow;
 
   private final TreeSet<KeyExtent> badExtents = new TreeSet<>();
+  private final HashSet<String> badServers = new HashSet<>();
   private final ReentrantReadWriteLock rwLock = new ReentrantReadWriteLock();
   private final Lock rLock = rwLock.readLock();
   private final Lock wLock = rwLock.writeLock();
@@ -446,16 +447,10 @@ public class TabletLocatorImpl extends TabletLocator {
 
   @Override
   public void invalidateCache(ClientContext context, String server) {
-    int invalidatedCount = 0;
 
     wLock.lock();
     try {
-      for (TabletLocation cacheEntry : metaCache.values()) {
-        if (cacheEntry.tablet_location.equals(server)) {
-          badExtents.add(cacheEntry.tablet_extent);
-          invalidatedCount++;
-        }
-      }
+      badServers.add(server);
     } finally {
       wLock.unlock();
     }
@@ -463,10 +458,8 @@ public class TabletLocatorImpl extends TabletLocator {
     lockChecker.invalidateCache(server);
 
     if (log.isTraceEnabled()) {
-      log.trace("invalidated {} cache entries  table={} server={}", 
invalidatedCount, tableId,
-          server);
+      log.trace("queued invalidation for table={} server={}", tableId, server);
     }
-
   }
 
   @Override
@@ -725,7 +718,7 @@ public class TabletLocatorImpl extends TabletLocator {
   private void processInvalidated(ClientContext context, LockCheckerSession 
lcSession)
       throws AccumuloSecurityException, AccumuloException, 
TableNotFoundException {
 
-    if (badExtents.isEmpty()) {
+    if (badExtents.isEmpty() && badServers.isEmpty()) {
       return;
     }
 
@@ -734,7 +727,7 @@ public class TabletLocatorImpl extends TabletLocator {
       if (!writeLockHeld) {
         rLock.unlock();
         wLock.lock();
-        if (badExtents.isEmpty()) {
+        if (badExtents.isEmpty() && badServers.isEmpty()) {
           return;
         }
       }
@@ -746,6 +739,26 @@ public class TabletLocatorImpl extends TabletLocator {
         removeOverlapping(metaCache, be);
       }
 
+      if (!badServers.isEmpty()) {
+        int removedCount = 0;
+        var locationIterator = metaCache.values().iterator();
+        while (locationIterator.hasNext()) {
+          TabletLocation cacheEntry = locationIterator.next();
+          if (badServers.contains(cacheEntry.tablet_location)) {
+            locationIterator.remove();
+            lookups.add(cacheEntry.tablet_extent.toMetaRange());
+            removedCount++;
+          }
+        }
+
+        if (log.isTraceEnabled()) {
+          log.trace("Invalidated {} cache entries for table {} related to 
servers {}", removedCount,
+              tableId, badServers);
+        }
+
+        badServers.clear();
+      }
+
       lookups = Range.mergeOverlapping(lookups);
 
       Map<String,Map<KeyExtent,List<Range>>> binnedRanges = new HashMap<>();