fix adn_device_ids_daily.sh

azkaban/adn/adn_device_ids_daily.sh

@@ -28,11 +28,12 @@ hadoop fs -rm -r ${OUTPUT_PATH}/*
 export HADOOP_CLIENT_OPTS="-Xmx2096m $HADOOP_CLIENT_OPTS"
 
 hadoop jar ../${JAR} mobvista.dmp.datasource.adn.mapreduce.AdnDeviceIdsMR \
-    -Dmapreduce.job.reduces=200 \
-    -Dmapreduce.map.memory.mb=4096 \
-    -Dmapreduce.map.java.opts=-Xmx2458m \
-    -Dmapreduce.reduce.memory.mb=4096 \
-    -Dmapreduce.reduce.java.opts=-Xmx2458m \
+    -Dmapreduce.job.reduces=1000 \
+    -Dmapred.child.java.opts=-Xmx4096m \
+    -Dmapreduce.map.memory.mb=8192 \
+    -Dmapreduce.map.java.opts=-Xmx4096m \
+    -Dmapreduce.reduce.memory.mb=8192 \
+    -Dmapreduce.reduce.java.opts=-Xmx4096m \
     -Dtask.date="$year-$month-$day" \
     -Dmapreduce.fileoutputcommitter.algorithm.version=2 \
     "$INPUT_ADN_DSP_PATH" "$OUTPUT_PATH" || exit 1

src/main/java/mobvista/dmp/datasource/adn/mapreduce/AdnDeviceIdsMR.java

@@ -126,12 +126,13 @@ public class AdnDeviceIdsMR extends Configured implements Tool {
         Configuration conf = getConf();
         conf.set("mapreduce.map.speculative", "true");
         conf.set("mapreduce.reduce.speculative", "true");
-        conf.set("mapreduce.task.io.sort.mb", "500");
-        conf.set("mapreduce.reduce.java.opts", "-Xmx1536m");
-        conf.set("mapreduce.reduce.memory.mb", "2048");
-        conf.set("mapreduce.reduce.shuffle.parallelcopies", "50");
+        conf.set("mapreduce.task.io.sort.mb", "1024");
+        conf.set("mapreduce.reduce.java.opts", "-Xmx4096m");
+        conf.set("mapreduce.reduce.memory.mb", "8192");
+        conf.set("mapreduce.reduce.shuffle.parallelcopies", "200");
         Job job = Job.getInstance(conf, "adn device ids job");
         job.setJarByClass(AdnDeviceIdsMR.class);
+        job.setNumReduceTasks(2000);
 
         FileInputFormat.addInputPath(job, new Path(args[0]));
         //FileInputFormat.addInputPath(job, new Path(args[1]));