Pair Machine|Processes, queue|cert, and sockets|hostname; fix .split auto-margins so cards fill half the row; trim Status prose and machine details; drop the page section index; note panel-docs on the roadmap. Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
+101
-36
@@ -62,10 +62,14 @@ type CPU struct {
|
||||
// nothing until it is true.
|
||||
Measured bool
|
||||
BusyPct float64
|
||||
Cores int
|
||||
// Cores and Threads are physical cores and logical CPUs (SMT). Both come
|
||||
// from /proc/cpuinfo when it is readable; otherwise Threads falls back to
|
||||
// the cpuN lines in /proc/stat and Cores matches Threads.
|
||||
Cores int
|
||||
Threads int
|
||||
// Load is the 1/5/15-minute load average, present when /proc/loadavg
|
||||
// could be read. Unlike BusyPct it needs no previous sample, so it is
|
||||
// there on the very first page load.
|
||||
// could be read. Unlike BusyPct it needs no previous sample. It is kept
|
||||
// for the sampler but is no longer shown on the Status card.
|
||||
Load [3]float64
|
||||
HasLoad bool
|
||||
Status Status
|
||||
@@ -176,7 +180,14 @@ type netCounters struct {
|
||||
func (m *MachineSampler) Sample() Machine {
|
||||
root := m.root()
|
||||
now := time.Now()
|
||||
cpuNow, cores, cpuErr := readCPUTimes(root)
|
||||
cpuNow, logical, cpuErr := readCPUTimes(root)
|
||||
cores, threads := readCPUTopology(root)
|
||||
if threads == 0 {
|
||||
threads = logical
|
||||
}
|
||||
if cores == 0 {
|
||||
cores = threads
|
||||
}
|
||||
netNow, netErr := readNetDev(root)
|
||||
|
||||
m.mu.Lock()
|
||||
@@ -198,7 +209,7 @@ func (m *MachineSampler) Sample() Machine {
|
||||
fresh := !prevAt.IsZero() && window > 0 && window <= machineSampleWindow
|
||||
|
||||
mach := Machine{
|
||||
CPU: cpuUsage(prevCPU, cpuNow, cores, readLoadAvg(root), fresh, cpuErr),
|
||||
CPU: cpuUsage(prevCPU, cpuNow, cores, threads, readLoadAvg(root), fresh, cpuErr),
|
||||
Memory: readMemory(root),
|
||||
Network: networkUsage(prevNet, netNow, window, fresh, netErr),
|
||||
}
|
||||
@@ -209,12 +220,11 @@ func (m *MachineSampler) Sample() Machine {
|
||||
return mach
|
||||
}
|
||||
|
||||
// cpuUsage grades the processor over the window. The load average is reported
|
||||
// alongside it because the two answer different questions — how busy the
|
||||
// processor was, and how many tasks were waiting for it — and a machine can
|
||||
// look idle while work queues up behind a slow disk.
|
||||
func cpuUsage(prev, cur cpuTimes, cores int, load [3]float64, fresh bool, err error) CPU {
|
||||
c := CPU{Cores: cores}
|
||||
// cpuUsage grades the processor over the window. Detail carries only the
|
||||
// core and thread counts — load average and busy prose stay out of the Status
|
||||
// card (see roadmap panel-docs for operator-facing explanation later).
|
||||
func cpuUsage(prev, cur cpuTimes, cores, threads int, load [3]float64, fresh bool, err error) CPU {
|
||||
c := CPU{Cores: cores, Threads: threads}
|
||||
if !isZeroLoad(load) {
|
||||
c.Load, c.HasLoad = load, true
|
||||
}
|
||||
@@ -223,9 +233,9 @@ func cpuUsage(prev, cur cpuTimes, cores int, load [3]float64, fresh bool, err er
|
||||
c.Detail = "The kernel's processor counters (/proc/stat) could not be read here."
|
||||
return c
|
||||
}
|
||||
c.Detail = c.contextText()
|
||||
if !fresh || cur.total <= prev.total {
|
||||
c.Status = StatusUnknown
|
||||
c.Detail = joinDetail("Measuring — this reading sets the baseline; the next refresh has the figure.", c.contextText())
|
||||
return c
|
||||
}
|
||||
|
||||
@@ -242,25 +252,24 @@ func cpuUsage(prev, cur cpuTimes, cores int, load [3]float64, fresh bool, err er
|
||||
|
||||
if c.BusyPct >= cpuWarnPct {
|
||||
c.Status = StatusWarn
|
||||
c.Detail = joinDetail(c.contextText(), "The processor is close to fully busy, which slows queue processing and every panel page.")
|
||||
} else {
|
||||
c.Status = StatusOK
|
||||
c.Detail = c.contextText()
|
||||
}
|
||||
return c
|
||||
}
|
||||
|
||||
// contextText is the CPU's supporting figures: what the percentage is a
|
||||
// percentage of, and how deep the run queue is.
|
||||
// contextText is the CPU detail column: physical cores and logical threads.
|
||||
func (c CPU) contextText() string {
|
||||
var parts []string
|
||||
if c.Cores > 0 {
|
||||
parts = append(parts, fmt.Sprintf("%d core(s)", c.Cores))
|
||||
switch {
|
||||
case c.Cores > 0 && c.Threads > 0:
|
||||
return fmt.Sprintf("%d cores · %d threads", c.Cores, c.Threads)
|
||||
case c.Threads > 0:
|
||||
return fmt.Sprintf("%d threads", c.Threads)
|
||||
case c.Cores > 0:
|
||||
return fmt.Sprintf("%d cores", c.Cores)
|
||||
default:
|
||||
return ""
|
||||
}
|
||||
if c.HasLoad {
|
||||
parts = append(parts, fmt.Sprintf("load average %.2f, %.2f, %.2f", c.Load[0], c.Load[1], c.Load[2]))
|
||||
}
|
||||
return strings.Join(parts, " · ")
|
||||
}
|
||||
|
||||
// readMemory reports main memory from /proc/meminfo. Used is derived from
|
||||
@@ -296,8 +305,7 @@ func readMemory(root string) Memory {
|
||||
m.SwapUsedBytes = m.SwapTotalBytes - swapFree
|
||||
}
|
||||
|
||||
detail := fmt.Sprintf("%s used of %s; %s available to new work.",
|
||||
humanBytes(m.UsedBytes), humanBytes(total), humanBytes(available))
|
||||
detail := fmt.Sprintf("%s used of %s.", humanBytes(m.UsedBytes), humanBytes(total))
|
||||
if m.SwapTotalBytes > 0 {
|
||||
detail += fmt.Sprintf(" Swap: %s of %s.", humanBytes(m.SwapUsedBytes), humanBytes(m.SwapTotalBytes))
|
||||
}
|
||||
@@ -418,6 +426,74 @@ func readCPUTimes(root string) (cpuTimes, int, error) {
|
||||
return times, cores, nil
|
||||
}
|
||||
|
||||
// readCPUTopology returns physical core and logical thread counts from
|
||||
// /proc/cpuinfo. Zeroes mean the file was missing or empty; the caller falls
|
||||
// back to the cpuN count from /proc/stat.
|
||||
func readCPUTopology(root string) (cores, threads int) {
|
||||
data, err := os.ReadFile(filepath.Join(root, "cpuinfo"))
|
||||
if err != nil {
|
||||
return 0, 0
|
||||
}
|
||||
type coreKey struct{ phys, core int }
|
||||
seen := map[coreKey]bool{}
|
||||
var (
|
||||
inCPU bool
|
||||
phys = -1
|
||||
core = -1
|
||||
idx int
|
||||
)
|
||||
flush := func() {
|
||||
if !inCPU {
|
||||
return
|
||||
}
|
||||
p, c := phys, core
|
||||
if p < 0 {
|
||||
p = idx
|
||||
}
|
||||
if c < 0 {
|
||||
c = idx
|
||||
}
|
||||
seen[coreKey{p, c}] = true
|
||||
inCPU, phys, core = false, -1, -1
|
||||
}
|
||||
for _, line := range strings.Split(string(data), "\n") {
|
||||
line = strings.TrimSpace(line)
|
||||
if line == "" {
|
||||
flush()
|
||||
continue
|
||||
}
|
||||
key, val, ok := strings.Cut(line, ":")
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
key = strings.TrimSpace(key)
|
||||
val = strings.TrimSpace(val)
|
||||
switch key {
|
||||
case "processor":
|
||||
flush()
|
||||
inCPU = true
|
||||
threads++
|
||||
idx = threads - 1
|
||||
if n, err := strconv.Atoi(val); err == nil {
|
||||
idx = n
|
||||
}
|
||||
case "physical id":
|
||||
if n, err := strconv.Atoi(val); err == nil {
|
||||
phys = n
|
||||
}
|
||||
case "core id":
|
||||
if n, err := strconv.Atoi(val); err == nil {
|
||||
core = n
|
||||
}
|
||||
}
|
||||
}
|
||||
flush()
|
||||
if threads == 0 {
|
||||
return 0, 0
|
||||
}
|
||||
return len(seen), threads
|
||||
}
|
||||
|
||||
// readLoadAvg reads the 1/5/15-minute load averages. A machine without
|
||||
// /proc/loadavg simply has none reported, so the failure is a zero value rather
|
||||
// than an error.
|
||||
@@ -544,14 +620,3 @@ func humanRate(perSec float64) string {
|
||||
}
|
||||
return humanBytes(uint64(perSec+0.5)) + "/s"
|
||||
}
|
||||
|
||||
// joinDetail joins the non-empty parts of a detail line.
|
||||
func joinDetail(parts ...string) string {
|
||||
var kept []string
|
||||
for _, p := range parts {
|
||||
if p != "" {
|
||||
kept = append(kept, p)
|
||||
}
|
||||
}
|
||||
return strings.Join(kept, " ")
|
||||
}
|
||||
|
||||
@@ -47,7 +47,7 @@ func TestCPUUsageOverTwoReadings(t *testing.T) {
|
||||
// 1000 ticks pass, 250 of them idle: 75% busy.
|
||||
prev := cpuTimes{total: 10000, idle: 8000}
|
||||
cur := cpuTimes{total: 11000, idle: 8250}
|
||||
got := cpuUsage(prev, cur, 4, [3]float64{0.5, 0.4, 0.3}, true, nil)
|
||||
got := cpuUsage(prev, cur, 4, 8, [3]float64{0.5, 0.4, 0.3}, true, nil)
|
||||
if !got.Measured {
|
||||
t.Fatalf("reading not marked measured: %+v", got)
|
||||
}
|
||||
@@ -57,16 +57,22 @@ func TestCPUUsageOverTwoReadings(t *testing.T) {
|
||||
if got.Status != StatusOK {
|
||||
t.Errorf("status = %q, want ok", got.Status)
|
||||
}
|
||||
if got.Cores != 4 || !got.HasLoad {
|
||||
t.Errorf("cores/load not reported: %+v", got)
|
||||
if got.Cores != 4 || got.Threads != 8 || !got.HasLoad {
|
||||
t.Errorf("cores/threads/load not reported: %+v", got)
|
||||
}
|
||||
if got.Detail != "4 cores · 8 threads" {
|
||||
t.Errorf("detail = %q, want cores and threads only", got.Detail)
|
||||
}
|
||||
}
|
||||
|
||||
func TestCPUUsageWarnsWhenFullyBusy(t *testing.T) {
|
||||
got := cpuUsage(cpuTimes{total: 10000, idle: 5000}, cpuTimes{total: 11000, idle: 5010}, 1, [3]float64{}, true, nil)
|
||||
got := cpuUsage(cpuTimes{total: 10000, idle: 5000}, cpuTimes{total: 11000, idle: 5010}, 1, 1, [3]float64{}, true, nil)
|
||||
if got.Status != StatusWarn {
|
||||
t.Errorf("99%% busy graded %q, want warn (%s)", got.Status, got.Detail)
|
||||
}
|
||||
if got.Detail != "1 cores · 1 threads" {
|
||||
t.Errorf("detail = %q, want topology only (no warn prose)", got.Detail)
|
||||
}
|
||||
if got.HasLoad {
|
||||
t.Error("a missing load average should not be reported as zeros")
|
||||
}
|
||||
@@ -84,17 +90,44 @@ func TestCPUUsageWithoutAUsableWindow(t *testing.T) {
|
||||
{"no previous reading", cpuTimes{}, cpuTimes{total: 11000, idle: 8250}, false},
|
||||
{"counters did not advance", cpuTimes{total: 11000, idle: 8250}, cpuTimes{total: 11000, idle: 8250}, true},
|
||||
} {
|
||||
got := cpuUsage(c.prev, c.cur, 2, [3]float64{}, c.fresh, nil)
|
||||
got := cpuUsage(c.prev, c.cur, 2, 2, [3]float64{}, c.fresh, nil)
|
||||
if got.Measured || got.Status != StatusUnknown {
|
||||
t.Errorf("%s: measured=%v status=%q, want unmeasured/unknown", c.name, got.Measured, got.Status)
|
||||
}
|
||||
}
|
||||
|
||||
if got := cpuUsage(cpuTimes{}, cpuTimes{}, 0, [3]float64{}, false, os.ErrNotExist); got.Status != StatusUnknown {
|
||||
if got := cpuUsage(cpuTimes{}, cpuTimes{}, 0, 0, [3]float64{}, false, os.ErrNotExist); got.Status != StatusUnknown {
|
||||
t.Errorf("unreadable /proc/stat: status %q, want unknown", got.Status)
|
||||
}
|
||||
}
|
||||
|
||||
func TestReadCPUTopology(t *testing.T) {
|
||||
dir := t.TempDir()
|
||||
const cpuinfo = `processor : 0
|
||||
physical id : 0
|
||||
core id : 0
|
||||
|
||||
processor : 1
|
||||
physical id : 0
|
||||
core id : 0
|
||||
|
||||
processor : 2
|
||||
physical id : 0
|
||||
core id : 1
|
||||
|
||||
processor : 3
|
||||
physical id : 0
|
||||
core id : 1
|
||||
`
|
||||
if err := os.WriteFile(filepath.Join(dir, "cpuinfo"), []byte(cpuinfo), 0o600); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
cores, threads := readCPUTopology(dir)
|
||||
if cores != 2 || threads != 4 {
|
||||
t.Errorf("topology = %d cores / %d threads, want 2/4", cores, threads)
|
||||
}
|
||||
}
|
||||
|
||||
func TestReadMemory(t *testing.T) {
|
||||
dir := fakeProc(t, "cpu 1 1 1 1 1 1 1 1\n", meminfoSample, "0.1 0.2 0.3 1/2 3\n", netdevSample)
|
||||
got := readMemory(dir)
|
||||
@@ -104,8 +137,7 @@ func TestReadMemory(t *testing.T) {
|
||||
if got.TotalBytes != 4*1024*1024*1024 {
|
||||
t.Errorf("total = %d bytes (%s), want 4 GiB", got.TotalBytes, got.TotalText())
|
||||
}
|
||||
// 4 GiB total, 2 GiB available to new work: half used, cache included in
|
||||
// what is available.
|
||||
// 4 GiB total, 2 GiB MemAvailable: half used (cache counted as available).
|
||||
if got.Percent() != 50 {
|
||||
t.Errorf("used = %.1f%% (%s), want 50%%", got.UsedPct, got.PctText())
|
||||
}
|
||||
@@ -221,8 +253,11 @@ func TestMachineSamplerNeedsTwoReadings(t *testing.T) {
|
||||
if !first.Memory.Measured {
|
||||
t.Error("memory is a level, not a rate: it must be reported on the first sample")
|
||||
}
|
||||
if first.CPU.Cores != 2 {
|
||||
t.Errorf("cores = %d, want 2", first.CPU.Cores)
|
||||
if first.CPU.Cores != 2 || first.CPU.Threads != 2 {
|
||||
t.Errorf("cores/threads = %d/%d, want 2/2", first.CPU.Cores, first.CPU.Threads)
|
||||
}
|
||||
if first.CPU.Detail != "2 cores · 2 threads" {
|
||||
t.Errorf("detail = %q, want topology only", first.CPU.Detail)
|
||||
}
|
||||
if !first.CPU.HasLoad || first.CPU.Load[0] != 0.42 {
|
||||
t.Errorf("load average not read: %+v", first.CPU.Load)
|
||||
|
||||
@@ -36,7 +36,7 @@ func CheckSocket(name, path string, required bool) Socket {
|
||||
default:
|
||||
s.Present = true
|
||||
s.Status = StatusOK
|
||||
s.Detail = "Listening."
|
||||
s.Detail = "Listening"
|
||||
}
|
||||
return s
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user