Merge branch 'cassandra-2.1.0' into cassandra-2.1

CHANGES.txt

@@ -10,6 +10,8 @@
 
 
 2.1.0-final
+ * Updated memtable_cleanup_threshold and memtable_flush_writers defaults 
+   (CASSANDRA-7551)
  * (Windows) fix startup when WMI memory query fails (CASSANDRA-7505)
  * Anti-compaction proceeds if any part of the repair failed (CASANDRA-7521)
  * Add missing table name to DROP INDEX responses and notifications (CASSANDRA-7539)

conf/cassandra.yaml

@@ -282,8 +282,13 @@ concurrent_counter_writes: 32
 # memtable_total_space_in_mb: 2048
 
 # Ratio of occupied non-flushing memtable size to total permitted size
-# that will trigger a flush of the largest memtable.
-memtable_cleanup_threshold: 0.4
+# that will trigger a flush of the largest memtable.  Lager mct will
+# mean larger flushes and hence less compaction, but also less concurrent
+# flush activity which can make it difficult to keep your disks fed
+# under heavy write load.
+#
+# memtable_cleanup_threshold defaults to 1 / (memtable_flush_writers + 1)
+#memtable_cleanup_threshold: 0.11
 
 # Specify the way Cassandra allocates and manages memtable memory.
 # Options are:
@@ -304,9 +309,14 @@ memtable_allocation_type: heap_buffers
 
 # This sets the amount of memtable flush writer threads.  These will
 # be blocked by disk io, and each one will hold a memtable in memory
-# while blocked. If your flush directory is backed by SSD, you may
-# want to increase this.
-memtable_flush_writers: 2
+# while blocked. 
+#
+# memtable_flush_writers defaults to the smaller of (number of disks,
+# number of cores), with a minimum of 2 and a maximum of 8.
+# 
+# If your data directories are backed by SSD, you should increase this
+# to the number of cores.
+#memtable_flush_writers: 8
 
 # A fixed memory pool size in MB for for SSTable index summaries. If left
 # empty, this will default to 5% of the heap size. If the memory usage of
@@ -511,6 +521,9 @@ batch_size_warn_threshold_in_kb: 5
 #
 # concurrent_compactors defaults to the smaller of (number of disks,
 # number of cores), with a minimum of 2 and a maximum of 8.
+# 
+# If your data directories are backed by SSD, you should increase this
+# to the number of cores.
 #concurrent_compactors: 1
 
 # Throttles compaction to the given total throughput across the entire

src/java/org/apache/cassandra/config/Config.java

@@ -90,12 +90,10 @@ public class Config
     @Deprecated
     public Integer concurrent_replicates = null;
 
-    // we don't want a lot of contention, but we also don't want to starve all other tables
-    // if a big one flushes. OS buffering should be able to minimize contention with 2 threads.
-    public int memtable_flush_writers = 2;
+    public Integer memtable_flush_writers = null;
     public Integer memtable_heap_space_in_mb;
     public Integer memtable_offheap_space_in_mb;
-    public float memtable_cleanup_threshold = 0.4f;
+    public Double memtable_cleanup_threshold = null;
 
     public Integer storage_port = 7000;
     public Integer ssl_storage_port = 7001;

src/java/org/apache/cassandra/config/DatabaseDescriptor.java

@@ -522,6 +522,12 @@ public class DatabaseDescriptor
         if (conf.commitlog_directory.equals(conf.saved_caches_directory))
             throw new ConfigurationException("saved_caches_directory must not be the same as the commitlog_directory");
 
+        if (conf.memtable_flush_writers == null)
+            conf.memtable_flush_writers = Math.min(8, Math.max(2, Math.min(FBUtilities.getAvailableProcessors(), conf.data_file_directories.length)));
+
+        if (conf.memtable_cleanup_threshold == null)
+            conf.memtable_cleanup_threshold = 1.0 / (1 + conf.memtable_flush_writers);
+
         if (conf.concurrent_compactors == null)
             conf.concurrent_compactors = Math.min(8, Math.max(2, Math.min(FBUtilities.getAvailableProcessors(), conf.data_file_directories.length)));