You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@ambari.apache.org by yu...@apache.org on 2016/02/12 19:30:59 UTC

[13/19] ambari git commit: AMBARI-14982 : Change AMS configs to deal with slow writes to WASB (avijayan)

AMBARI-14982 : Change AMS configs to deal with slow writes to WASB (avijayan)


Project: http://git-wip-us.apache.org/repos/asf/ambari/repo
Commit: http://git-wip-us.apache.org/repos/asf/ambari/commit/9120682e
Tree: http://git-wip-us.apache.org/repos/asf/ambari/tree/9120682e
Diff: http://git-wip-us.apache.org/repos/asf/ambari/diff/9120682e

Branch: refs/heads/2.2.1-maint
Commit: 9120682ed5de26c6163c6bc07935782fb645ff32
Parents: 984d9e9
Author: Aravindan Vijayan <av...@hortonworks.com>
Authored: Thu Feb 11 10:16:48 2016 -0800
Committer: Aravindan Vijayan <av...@hortonworks.com>
Committed: Thu Feb 11 10:16:48 2016 -0800

----------------------------------------------------------------------
 .../timeline/HBaseTimelineMetricStore.java      | 18 +++++-----
 .../metrics/timeline/PhoenixHBaseAccessor.java  | 21 ++++++-----
 .../timeline/TimelineMetricConfiguration.java   | 10 ++++++
 .../0.1.0/configuration/ams-site.xml            | 37 +++++++++++++++-----
 .../stacks/HDP/2.0.6/services/stack_advisor.py  |  5 ++-
 .../stacks/2.2/common/test_stack_advisor.py     |  5 ++-
 6 files changed, 69 insertions(+), 27 deletions(-)
----------------------------------------------------------------------


http://git-wip-us.apache.org/repos/asf/ambari/blob/9120682e/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java
----------------------------------------------------------------------
diff --git a/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java b/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java
index c4e946a..8678071 100644
--- a/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java
+++ b/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/HBaseTimelineMetricStore.java
@@ -142,14 +142,16 @@ public class HBaseTimelineMetricStore extends AbstractService implements Timelin
         aggregatorDailyThread.start();
       }
 
-      int initDelay = configuration.getTimelineMetricsServiceWatcherInitDelay();
-      int delay = configuration.getTimelineMetricsServiceWatcherDelay();
-      // Start the watchdog
-      executorService.scheduleWithFixedDelay(
-        new TimelineMetricStoreWatcher(this, configuration), initDelay, delay,
-        TimeUnit.SECONDS);
-      LOG.info("Started watchdog for timeline metrics store with initial " +
-        "delay = " + initDelay + ", delay = " + delay);
+      if (!configuration.isTimelineMetricsServiceWatcherDisabled()) {
+        int initDelay = configuration.getTimelineMetricsServiceWatcherInitDelay();
+        int delay = configuration.getTimelineMetricsServiceWatcherDelay();
+        // Start the watchdog
+        executorService.scheduleWithFixedDelay(
+          new TimelineMetricStoreWatcher(this, configuration), initDelay, delay,
+          TimeUnit.SECONDS);
+        LOG.info("Started watchdog for timeline metrics store with initial " +
+          "delay = " + initDelay + ", delay = " + delay);
+      }
 
       isInitialized = true;
     }

http://git-wip-us.apache.org/repos/asf/ambari/blob/9120682e/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java
----------------------------------------------------------------------
diff --git a/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java b/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java
index 4dc1a95..60ac06f 100644
--- a/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java
+++ b/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/PhoenixHBaseAccessor.java
@@ -244,14 +244,14 @@ public class PhoenixHBaseAccessor {
 
     String encoding = metricsConf.get(HBASE_ENCODING_SCHEME, DEFAULT_ENCODING);
     String compression = metricsConf.get(HBASE_COMPRESSION_SCHEME, DEFAULT_TABLE_COMPRESSION);
-    String precisionTtl = metricsConf.get(PRECISION_TABLE_TTL, "86400");           //1 day
-    String hostMinTtl = metricsConf.get(HOST_MINUTE_TABLE_TTL, "604800");          //7 days
-    String hostHourTtl = metricsConf.get(HOST_HOUR_TABLE_TTL, "2592000");          //30 days
-    String hostDailyTtl = metricsConf.get(HOST_DAILY_TABLE_TTL, "31536000");       //1 year
-    String clusterSecTtl = metricsConf.get(CLUSTER_SECOND_TABLE_TTL, "2592000");     //7 days
-    String clusterMinTtl = metricsConf.get(CLUSTER_MINUTE_TABLE_TTL, "7776000");   //30 days
-    String clusterHourTtl = metricsConf.get(CLUSTER_HOUR_TABLE_TTL, "31536000");   //1 year
-    String clusterDailyTtl = metricsConf.get(CLUSTER_DAILY_TABLE_TTL, "63072000"); //2 years
+    String precisionTtl = getDaysInSeconds(metricsConf.get(PRECISION_TABLE_TTL, "1"));           //1 day
+    String hostMinTtl = getDaysInSeconds(metricsConf.get(HOST_MINUTE_TABLE_TTL, "7"));           //7 days
+    String hostHourTtl = getDaysInSeconds(metricsConf.get(HOST_HOUR_TABLE_TTL, "30"));           //30 days
+    String hostDailyTtl = getDaysInSeconds(metricsConf.get(HOST_DAILY_TABLE_TTL, "365"));        //1 year
+    String clusterSecTtl = getDaysInSeconds(metricsConf.get(CLUSTER_SECOND_TABLE_TTL, "7"));     //7 days
+    String clusterMinTtl = getDaysInSeconds(metricsConf.get(CLUSTER_MINUTE_TABLE_TTL, "30"));    //30 days
+    String clusterHourTtl = getDaysInSeconds(metricsConf.get(CLUSTER_HOUR_TABLE_TTL, "365"));    //1 year
+    String clusterDailyTtl = getDaysInSeconds(metricsConf.get(CLUSTER_DAILY_TABLE_TTL, "730"));  //2 years
 
     try {
       LOG.info("Initializing metrics schema...");
@@ -1023,4 +1023,9 @@ public class PhoenixHBaseAccessor {
         "thread = " + Thread.currentThread().getName());
     }
   }
+
+  private String getDaysInSeconds(String daysString) {
+    double days = Double.valueOf(daysString.trim());
+    return String.valueOf((int)(days*86400));
+  }
 }

http://git-wip-us.apache.org/repos/asf/ambari/blob/9120682e/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java
----------------------------------------------------------------------
diff --git a/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java b/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java
index 6311f9f..c0ab6f1 100644
--- a/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java
+++ b/ambari-metrics/ambari-metrics-timelineservice/src/main/java/org/apache/hadoop/yarn/server/applicationhistoryservice/metrics/timeline/TimelineMetricConfiguration.java
@@ -181,6 +181,9 @@ public class TimelineMetricConfiguration {
   public static final String HANDLER_THREAD_COUNT =
     "timeline.metrics.service.handler.thread.count";
 
+  public static final String WATCHER_DISABLED =
+    "timeline.metrics.service.watcher.disabled";
+
   public static final String WATCHER_INITIAL_DELAY =
     "timeline.metrics.service.watcher.initial.delay";
 
@@ -261,6 +264,13 @@ public class TimelineMetricConfiguration {
     return 20;
   }
 
+  public boolean isTimelineMetricsServiceWatcherDisabled() {
+    if (metricsConf != null) {
+      return Boolean.parseBoolean(metricsConf.get(WATCHER_DISABLED, "false"));
+    }
+    return false;
+  }
+
   public int getTimelineMetricsServiceWatcherInitDelay() {
     if (metricsConf != null) {
       return Integer.parseInt(metricsConf.get(WATCHER_INITIAL_DELAY, "600"));

http://git-wip-us.apache.org/repos/asf/ambari/blob/9120682e/ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml
----------------------------------------------------------------------
diff --git a/ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml b/ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml
index 8173743..5f66063 100644
--- a/ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml
+++ b/ambari-server/src/main/resources/common-services/AMBARI_METRICS/0.1.0/configuration/ams-site.xml
@@ -283,59 +283,65 @@
   </property>
   <property>
     <name>timeline.metrics.host.aggregator.daily.ttl</name>
-    <value>31536000</value>
+    <value>365</value>
     <description>
       Host based daily resolution data purge interval. Default is 1 year.
     </description>
   </property>
   <property>
     <name>timeline.metrics.host.aggregator.hourly.ttl</name>
-    <value>2592000</value>
+    <value>30</value>
     <description>
       Host based hourly resolution data purge interval. Default is 30 days.
     </description>
   </property>
   <property>
     <name>timeline.metrics.host.aggregator.minute.ttl</name>
-    <value>604800</value>
+    <value>7</value>
     <description>
       Host based minute resolution data purge interval. Default is 7 days.
     </description>
   </property>
   <property>
     <name>timeline.metrics.cluster.aggregator.second.ttl</name>
-    <value>2592000</value>
+    <value>7</value>
     <description>
       Cluster wide second resolution data purge interval. Default is 7 days.
     </description>
   </property>
   <property>
     <name>timeline.metrics.cluster.aggregator.minute.ttl</name>
-    <value>7776000</value>
+    <value>30</value>
     <description>
       Cluster wide minute resolution data purge interval. Default is 30 days.
     </description>
   </property>
   <property>
     <name>timeline.metrics.cluster.aggregator.hourly.ttl</name>
-    <value>31536000</value>
+    <value>365</value>
     <description>
       Cluster wide hourly resolution data purge interval. Default is 1 year.
     </description>
   </property>
   <property>
     <name>timeline.metrics.cluster.aggregator.daily.ttl</name>
-    <value>63072000</value>
+    <value>730</value>
     <description>
       Cluster wide daily resolution data purge interval. Default is 2 years.
     </description>
   </property>
   <property>
     <name>timeline.metrics.host.aggregator.ttl</name>
-    <value>86400</value>
+    <value>1</value>
     <description>
-      1 minute resolution data purge interval. Default is 1 day.
+      1 minute resolution data purge interval. Default is 1 day for embedded mode and 7 days for Distributed mode.
     </description>
+    <depends-on>
+      <property>
+        <type>ams-site</type>
+        <name>timeline.metrics.service.operation.mode</name>
+      </property>
+    </depends-on>
   </property>
   <property>
     <name>timeline.metrics.hbase.data.block.encoding</name>
@@ -492,6 +498,19 @@
     </description>
   </property>
   <property>
+    <name>timeline.metrics.service.watcher.disabled</name>
+    <value>false</value>
+    <description>
+      Disable Timeline Metric Store watcher thread. Disabled by default in AMS distributed mode.
+    </description>
+    <depends-on>
+      <property>
+        <type>ams-site</type>
+        <name>timeline.metrics.service.operation.mode</name>
+      </property>
+    </depends-on>
+  </property>
+  <property>
     <name>timeline.metrics.service.watcher.initial.delay</name>
     <value>600</value>
     <description>

http://git-wip-us.apache.org/repos/asf/ambari/blob/9120682e/ambari-server/src/main/resources/stacks/HDP/2.0.6/services/stack_advisor.py
----------------------------------------------------------------------
diff --git a/ambari-server/src/main/resources/stacks/HDP/2.0.6/services/stack_advisor.py b/ambari-server/src/main/resources/stacks/HDP/2.0.6/services/stack_advisor.py
index e8df97b..07cc73b 100644
--- a/ambari-server/src/main/resources/stacks/HDP/2.0.6/services/stack_advisor.py
+++ b/ambari-server/src/main/resources/stacks/HDP/2.0.6/services/stack_advisor.py
@@ -470,8 +470,12 @@ class HDP206StackAdvisor(DefaultStackAdvisor):
         operatingMode = services["configurations"]["ams-site"]["properties"]["timeline.metrics.service.operation.mode"]
 
     if operatingMode == "distributed":
+      putAmsSiteProperty("timeline.metrics.service.watcher.disabled", 'true')
+      putAmsSiteProperty("timeline.metrics.host.aggregator.ttl", 7)
       putAmsHbaseSiteProperty("hbase.cluster.distributed", 'true')
     else:
+      putAmsSiteProperty("timeline.metrics.service.watcher.disabled", 'false')
+      putAmsSiteProperty("timeline.metrics.host.aggregator.ttl", 1)
       putAmsHbaseSiteProperty("hbase.cluster.distributed", 'false')
 
     rootDir = "file:///var/lib/ambari-metrics-collector/hbase"
@@ -527,7 +531,6 @@ class HDP206StackAdvisor(DefaultStackAdvisor):
     putAmsHbaseSiteProperty("hbase.hregion.memstore.flush.size", 134217728)
     putAmsHbaseSiteProperty("hbase.regionserver.global.memstore.upperLimit", 0.35)
     putAmsHbaseSiteProperty("hbase.regionserver.global.memstore.lowerLimit", 0.3)
-    putAmsSiteProperty("timeline.metrics.host.aggregator.ttl", 86400)
 
     if len(amsCollectorHosts) > 1:
       pass

http://git-wip-us.apache.org/repos/asf/ambari/blob/9120682e/ambari-server/src/test/python/stacks/2.2/common/test_stack_advisor.py
----------------------------------------------------------------------
diff --git a/ambari-server/src/test/python/stacks/2.2/common/test_stack_advisor.py b/ambari-server/src/test/python/stacks/2.2/common/test_stack_advisor.py
index 86d560a..6984e0e 100644
--- a/ambari-server/src/test/python/stacks/2.2/common/test_stack_advisor.py
+++ b/ambari-server/src/test/python/stacks/2.2/common/test_stack_advisor.py
@@ -2011,7 +2011,8 @@ class TestHDP22StackAdvisor(TestCase):
         "properties": {
           "timeline.metrics.cluster.aggregate.splitpoints": " ",
           "timeline.metrics.host.aggregate.splitpoints": " ",
-          "timeline.metrics.host.aggregator.ttl": "86400"
+          "timeline.metrics.host.aggregator.ttl": "1",
+          'timeline.metrics.service.watcher.disabled': 'false'
         }
       }
     }
@@ -2189,6 +2190,8 @@ class TestHDP22StackAdvisor(TestCase):
     expected['ams-hbase-env']['properties']['hbase_regionserver_heapsize'] = '512'
     expected["ams-hbase-env"]['properties']['hbase_master_xmn_size'] = '102'
     expected['ams-hbase-env']['properties']['regionserver_xmn_size'] = '384'
+    expected['ams-site']['properties']['timeline.metrics.host.aggregator.ttl'] = '7'
+    expected['ams-site']['properties']['timeline.metrics.service.watcher.disabled'] = 'true'
     self.stackAdvisor.recommendAmsConfigurations(configurations, clusterData, services, hosts)
     self.assertEquals(configurations, expected)