= system-metrics: fix ProcessCpuMetrics usage in order to show the percent usage shared in the numbers of cpu cores

1 files changed, 33 insertions, 17 deletions

diff --git a/kamon-system-metrics/src/main/scala/kamon/system/sigar/ProcessCpuMetrics.scala b/kamon-system-metrics/src/main/scala/kamon/system/sigar/ProcessCpuMetrics.scala
index 4f197c8f..436fadda 100644
--- a/kamon-system-metrics/src/main/scala/kamon/system/sigar/ProcessCpuMetrics.scala
+++ b/kamon-system-metrics/src/main/scala/kamon/system/sigar/ProcessCpuMetrics.scala
@@ -25,31 +25,47 @@ class ProcessCpuMetrics(sigar: Sigar, instrumentFactory: InstrumentFactory) exte
   val processSystemCpu = histogram("process-system-cpu")
   val processTotalCpu = histogram("process-cpu")
 
-  var lastProcCpu: Option[ProcCpu] = None
+  val pid = sigar.getPid
+  val totalCores = sigar.getCpuInfoList.headOption.map(_.getTotalCores.toLong).getOrElse(1L)
 
+  var lastProcCpu: ProcCpu = sigar.getProcCpu(pid)
+  var currentLoad: Long = 0
+
+  /**
+   * While CPU usage time updates not very often, We have introduced a simple heuristic, that supposes that the load is the same as previous,
+   * while CPU usage time doesn't update. But supposing that it could be zero load for a process for some time,
+   * We used an arbitrary duration of 2000 milliseconds, after which the same CPU usage time value become legal, and it is supposed that the load is really zero.
+   *
+   * @see [[http://stackoverflow.com/questions/19323364/using-sigar-api-to-get-jvm-cpu-usage "Stack Overflow: Using Sigar API to get JVM Cpu usage"]]
+   */
   def update(): Unit = {
-    val pid = sigar.getPid
-    val procCpu = sigar.getProcCpu(pid)
-
-    lastProcCpu.map { last ⇒
-      val timeDiff = procCpu.getLastTime - last.getLastTime
-      if (timeDiff > 0) {
-        val userPercent = (((procCpu.getUser - last.getUser) / timeDiff.toDouble) * 100).toLong
-        val systemPercent = (((procCpu.getSys - last.getSys) / timeDiff.toDouble) * 100).toLong
-
-        processUserCpu.record(userPercent)
-        processSystemCpu.record(systemPercent)
-        processTotalCpu.record(userPercent + systemPercent)
-      }
-    }
+    val currentProcCpu = sigar.getProcCpu(pid)
+    val totalDiff = currentProcCpu.getTotal - lastProcCpu.getTotal
+    val userDiff = currentProcCpu.getUser - lastProcCpu.getUser
+    val systemDiff = currentProcCpu.getSys - lastProcCpu.getSys
+    val timeDiff = currentProcCpu.getLastTime - lastProcCpu.getLastTime
+
+    def percentUsage(delta: Long): Long = 100 * delta / timeDiff / totalCores
 
-    lastProcCpu = Some(procCpu)
+    if (totalDiff == 0) {
+      if (timeDiff > 2000) currentLoad = 0
+      if (currentLoad == 0) lastProcCpu = currentProcCpu
+    } else {
+      val totalPercent = percentUsage(totalDiff)
+      val userPercent = percentUsage(userDiff)
+      val systemPercent = percentUsage(systemDiff)
 
+      processUserCpu.record(userPercent)
+      processSystemCpu.record(systemPercent)
+      processTotalCpu.record(userPercent + systemPercent)
+
+      currentLoad = totalPercent
+      lastProcCpu = currentProcCpu
+    }
   }
 }
 
 object ProcessCpuMetrics extends SigarMetricRecorderCompanion("process-cpu") {
-
   def apply(sigar: Sigar, instrumentFactory: InstrumentFactory): ProcessCpuMetrics =
     new ProcessCpuMetrics(sigar, instrumentFactory)
 }