From 2d65dc418c88d8f41ec4af35e90001255eda4d56 Mon Sep 17 00:00:00 2001
From: Tim Gross <tgross@hashicorp.com>
Date: Tue, 24 Oct 2023 10:37:46 -0400
Subject: [PATCH] metrics: prevent negative counter from iowait decrease
 (#18849)

The iowait metric obtained from `/proc/stat` can under some circumstances
decrease. The relevant condition is when an interrupt arrives on a different
core than the one that gets woken up for the IO, and a particular counter in the
kernel for that core gets interrupted. This is documented in the man page for
the `proc(5)` pseudo-filesystem, and considered an unfortunate behavior that
can't be changed for the sake of ABI compatibility.

In Nomad, we get the current "busy" time (everything except for idle) and
compare it to the previous busy time to get the counter incremeent. If the
iowait counter decreases and the idle counter increases more than the increase
in the total busy time, we can get a negative total. This previously caused a
panic in our metrics collection (see #15861) but that is being prevented by
reporting an error message.

Fix the bug by putting a zero floor on the values we return from the host CPU
stats calculator.

Backport-of: #18835
---
 .changelog/18835.txt      |  3 +++
 client/stats/host.go      | 10 ++++-----
 client/stats/host_test.go | 43 ++++++++++++++++++++++++++++-----------
 3 files changed, 39 insertions(+), 17 deletions(-)
 create mode 100644 .changelog/18835.txt

diff --git a/.changelog/18835.txt b/.changelog/18835.txt
new file mode 100644
index 000000000..99d4ae17e
--- /dev/null
+++ b/.changelog/18835.txt
@@ -0,0 +1,3 @@
+```release-note:bug
+metrics: Fixed a bug where CPU counters could report errors for negative values
+```
diff --git a/client/stats/host.go b/client/stats/host.go
index bd6edb32d..a28e2716e 100644
--- a/client/stats/host.go
+++ b/client/stats/host.go
@@ -277,7 +277,7 @@ func (h *HostCpuStatsCalculator) Calculate(times cpu.TimesStat) (idle float64, u
 	currentIdle := times.Idle
 	currentUser := times.User
 	currentSystem := times.System
-	currentTotal := times.Total()
+	currentTotal := times.Total() // this is Idle + currentBusy
 	currentBusy := times.User + times.System + times.Nice + times.Iowait + times.Irq +
 		times.Softirq + times.Steal + times.Guest + times.GuestNice
 
@@ -288,16 +288,16 @@ func (h *HostCpuStatsCalculator) Calculate(times cpu.TimesStat) (idle float64, u
 	total = ((currentBusy - h.prevBusy) / deltaTotal) * 100
 
 	// Protect against any invalid values
-	if math.IsNaN(idle) || math.IsInf(idle, 0) {
+	if math.IsNaN(idle) || math.IsInf(idle, 0) || idle < 0.0 {
 		idle = 100.0
 	}
-	if math.IsNaN(user) || math.IsInf(user, 0) {
+	if math.IsNaN(user) || math.IsInf(user, 0) || user < 0.0 {
 		user = 0.0
 	}
-	if math.IsNaN(system) || math.IsInf(system, 0) {
+	if math.IsNaN(system) || math.IsInf(system, 0) || system < 0.0 {
 		system = 0.0
 	}
-	if math.IsNaN(total) || math.IsInf(total, 0) {
+	if math.IsNaN(total) || math.IsInf(total, 0) || total < 0 {
 		total = 0.0
 	}
 
diff --git a/client/stats/host_test.go b/client/stats/host_test.go
index 3d939cd10..4aada7ddf 100644
--- a/client/stats/host_test.go
+++ b/client/stats/host_test.go
@@ -7,6 +7,7 @@ import (
 	"testing"
 
 	"github.com/shirou/gopsutil/v3/cpu"
+	"github.com/shoenig/test/must"
 )
 
 func TestHostCpuStatsCalculator_Nan(t *testing.T) {
@@ -20,16 +21,34 @@ func TestHostCpuStatsCalculator_Nan(t *testing.T) {
 	calculator.Calculate(times)
 	idle, user, system, total := calculator.Calculate(times)
 
-	if idle != 100.0 {
-		t.Errorf("idle: Expected: %f, Got %f", 100.0, idle)
-	}
-	if user != 0.0 {
-		t.Errorf("user: Expected: %f, Got %f", 0.0, user)
-	}
-	if system != 0.0 {
-		t.Errorf("system: Expected: %f, Got %f", 0.0, system)
-	}
-	if total != 0.0 {
-		t.Errorf("total: Expected: %f, Got %f", 0.0, total)
-	}
+	must.Eq(t, 100.0, idle, must.Sprint("unexpected idle stats"))
+	must.Eq(t, 0.0, user, must.Sprint("unexpected user stats"))
+	must.Eq(t, 0.0, system, must.Sprint("unexpected system stats"))
+	must.Eq(t, 0.0, total, must.Sprint("unexpected total stats"))
+}
+
+func TestHostCpuStatsCalculator_DecreasedIOWait(t *testing.T) {
+	times := cpu.TimesStat{
+		CPU:    "cpu0",
+		User:   20000,
+		Nice:   100,
+		System: 9000,
+		Idle:   370000,
+		Iowait: 700,
+	}
+
+	calculator := NewHostCpuStatsCalculator()
+	calculator.Calculate(times)
+
+	times = cpu.TimesStat{
+		CPU:    "cpu0",
+		User:   20000,
+		Nice:   100,
+		System: 9000,
+		Idle:   380000,
+		Iowait: 600,
+	}
+
+	_, _, _, total := calculator.Calculate(times)
+	must.GreaterEq(t, 0.0, total, must.Sprint("total must never be negative"))
 }