From d210213fc9f5375a6c710b790196dc9ba33e2952 Mon Sep 17 00:00:00 2001
From: Chris Earle <pickypg@users.noreply.github.com>
Date: Fri, 30 Dec 2016 18:55:51 -0500
Subject: [PATCH] [Monitoring] Use Cgroup Data from _nodes/stats
 (elastic/elasticsearch#4484)

* [Monitoring] Use Cgroup Data from _nodes/stats

This adds Cgroup data from _nodes/stats.

- This also adds 5m and 15m from system load, which are sent from ES, but
  were not recorded. Kibana does record/report those values though.

Original commit: elastic/x-pack-elasticsearch@8e04452c60bf35721020eb9fc6513e2231b6793e
---
 .../resolver/node/NodeStatsResolver.java      | 11 ++++
 .../src/main/resources/monitoring-es.json     | 54 ++++++++++++++++++-
 .../resolver/node/NodeStatsResolverTests.java |  7 ++-
 3 files changed, 70 insertions(+), 2 deletions(-)

diff --git a/elasticsearch/src/main/java/org/elasticsearch/xpack/monitoring/resolver/node/NodeStatsResolver.java b/elasticsearch/src/main/java/org/elasticsearch/xpack/monitoring/resolver/node/NodeStatsResolver.java
index 47ab958d989..4da2334382f 100644
--- a/elasticsearch/src/main/java/org/elasticsearch/xpack/monitoring/resolver/node/NodeStatsResolver.java
+++ b/elasticsearch/src/main/java/org/elasticsearch/xpack/monitoring/resolver/node/NodeStatsResolver.java
@@ -66,6 +66,8 @@ public class NodeStatsResolver extends MonitoringIndexNameResolver.Timestamped<N
             "node_stats.fs.total.free_in_bytes",
             "node_stats.fs.total.available_in_bytes",
             "node_stats.os.cpu.load_average.1m",
+            "node_stats.os.cpu.load_average.5m",
+            "node_stats.os.cpu.load_average.15m",
             "node_stats.process.cpu.percent",
             "node_stats.process.max_file_descriptors",
             "node_stats.process.open_file_descriptors",
@@ -99,6 +101,15 @@ public class NodeStatsResolver extends MonitoringIndexNameResolver.Timestamped<N
             "node_stats.thread_pool.watcher.threads",
             "node_stats.thread_pool.watcher.queue",
             "node_stats.thread_pool.watcher.rejected",
+            // Cgroup data (generally Linux only and only sometimes on those systems)
+            "node_stats.os.cgroup.cpuacct.control_group",
+            "node_stats.os.cgroup.cpuacct.usage_nanos",
+            "node_stats.os.cgroup.cpu.control_group",
+            "node_stats.os.cgroup.cpu.cfs_period_micros",
+            "node_stats.os.cgroup.cpu.cfs_quota_micros",
+            "node_stats.os.cgroup.cpu.stat.number_of_elapsed_periods",
+            "node_stats.os.cgroup.cpu.stat.number_of_times_throttled",
+            "node_stats.os.cgroup.cpu.stat.time_throttled_nanos",
             // Linux Only (at least for now)
             // Disk Info
             "node_stats.fs.data.spins",
diff --git a/elasticsearch/src/main/resources/monitoring-es.json b/elasticsearch/src/main/resources/monitoring-es.json
index 3c434831ad6..22b334e4d62 100644
--- a/elasticsearch/src/main/resources/monitoring-es.json
+++ b/elasticsearch/src/main/resources/monitoring-es.json
@@ -581,7 +581,59 @@
               }
             },
             "os": {
-              "type": "object"
+              "properties": {
+                "cgroup": {
+                  "properties": {
+                    "cpuacct": {
+                      "properties": {
+                        "control_group": {
+                          "type": "keyword"
+                        },
+                        "usage_nanos": {
+                          "type": "long"
+                        }
+                      }
+                    },
+                    "cpu": {
+                      "properties": {
+                        "control_group": {
+                          "type": "keyword"
+                        },
+                        "stat": {
+                          "properties": {
+                            "number_of_elapsed_periods": {
+                              "type": "long"
+                            },
+                            "number_of_times_throttled": {
+                              "type": "long"
+                            },
+                            "time_throttled_nanos": {
+                              "type": "long"
+                            }
+                          }
+                        }
+                      }
+                    }
+                  }
+                },
+                "cpu": {
+                  "properties": {
+                    "load_average": {
+                      "properties": {
+                        "1m": {
+                          "type": "half_float"
+                        },
+                        "5m": {
+                          "type": "half_float"
+                        },
+                        "15m": {
+                          "type": "half_float"
+                        }
+                      }
+                    }
+                  }
+                }
+              }
             },
             "process": {
               "type": "object"
diff --git a/elasticsearch/src/test/java/org/elasticsearch/xpack/monitoring/resolver/node/NodeStatsResolverTests.java b/elasticsearch/src/test/java/org/elasticsearch/xpack/monitoring/resolver/node/NodeStatsResolverTests.java
index d67d26e509b..a30fad4b1c5 100644
--- a/elasticsearch/src/test/java/org/elasticsearch/xpack/monitoring/resolver/node/NodeStatsResolverTests.java
+++ b/elasticsearch/src/test/java/org/elasticsearch/xpack/monitoring/resolver/node/NodeStatsResolverTests.java
@@ -82,7 +82,7 @@ public class NodeStatsResolverTests extends MonitoringIndexNameResolverTestCase<
             return;
         }
 
-        // we only report IoStats and spins on Linux
+        // we only report IoStats on Linux
         if (Constants.LINUX == false) {
             if (field.startsWith("node_stats.fs.io_stats")) {
                 return;
@@ -94,6 +94,11 @@ public class NodeStatsResolverTests extends MonitoringIndexNameResolverTestCase<
             return;
         }
 
+        // cgroups can be null, and it's only reported on Linux
+        if (field.startsWith("node_stats.os.cgroup")) {
+            return;
+        }
+
         super.assertSourceField(field, sourceFields);
     }