Settings: one window for how far back every graph reaches, 1 minute to 24 hours
Each machine keeps 300 points per curve whatever the window, so memory doesn't follow it; a wider window means wider points, each the average of its bucket rather than one instant out of it. The hub passes the setting to every machine it manages. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
885b050d02
commit
04dd98e2cd
@@ -27,12 +27,21 @@ type statusSensors struct {
|
||||
cpuPrev map[string]cpuMark // container ID → CPU time at the previous read
|
||||
|
||||
hist map[string]*sensors.Series // numeric sensors only, statusHistory points each
|
||||
acc map[string]*bucket // the point being built for each, when the window is wider than the read rate
|
||||
}
|
||||
|
||||
// statusHistory is how many readings each status sensor keeps: at 15 s that is over an
|
||||
// hour of container memory; at 10 min, two days of drive wear.
|
||||
const statusHistory = 300
|
||||
|
||||
// A point being averaged, the same idea as the sampler's: these sources are read every 15 to
|
||||
// 300 seconds, so a wide window keeps one point per stride rather than one per read.
|
||||
type bucket struct {
|
||||
start int64
|
||||
sum float64
|
||||
n float64
|
||||
}
|
||||
|
||||
const (
|
||||
servicesEvery = 30 * time.Second
|
||||
containersEvery = 15 * time.Second
|
||||
@@ -42,7 +51,7 @@ const (
|
||||
)
|
||||
|
||||
func newStatusSensors(s *Server) *statusSensors {
|
||||
p := &statusSensors{s: s, by: map[string][]sensors.Sensor{}, hist: map[string]*sensors.Series{}}
|
||||
p := &statusSensors{s: s, by: map[string][]sensors.Sensor{}, hist: map[string]*sensors.Series{}, acc: map[string]*bucket{}}
|
||||
go p.loop("services", servicesEvery, p.services)
|
||||
go p.loop("containers", containersEvery, p.containers)
|
||||
go p.loop("updates", updatesEvery, p.updates)
|
||||
@@ -58,6 +67,8 @@ func (p *statusSensors) loop(name string, every time.Duration, read func(context
|
||||
cancel()
|
||||
p.mu.Lock()
|
||||
p.by[name] = got
|
||||
window := p.s.metrics.Window()
|
||||
stride := window.Milliseconds() / statusHistory
|
||||
live := map[string]bool{}
|
||||
for _, x := range got {
|
||||
v, ok := x.Value.(float64)
|
||||
@@ -70,14 +81,32 @@ func (p *statusSensors) loop(name string, every time.Duration, read func(context
|
||||
h = &sensors.Series{ID: x.ID, Unit: x.Unit}
|
||||
p.hist[x.ID] = h
|
||||
}
|
||||
h.T, h.V = append(h.T, x.T), append(h.V, v)
|
||||
b := p.acc[x.ID]
|
||||
if b == nil {
|
||||
b = &bucket{start: x.T}
|
||||
p.acc[x.ID] = b
|
||||
}
|
||||
b.sum, b.n = b.sum+v, b.n+1
|
||||
if x.T-b.start < stride && b.n > 1 {
|
||||
continue // still filling this point
|
||||
}
|
||||
h.T, h.V = append(h.T, x.T), append(h.V, b.sum/b.n)
|
||||
*b = bucket{start: x.T + 1}
|
||||
if len(h.T) > statusHistory {
|
||||
h.T, h.V = h.T[len(h.T)-statusHistory:], h.V[len(h.V)-statusHistory:]
|
||||
}
|
||||
if cut := x.T - window.Milliseconds(); h.T[0] < cut { // and nothing older than the window
|
||||
i := 0
|
||||
for i < len(h.T) && h.T[i] < cut {
|
||||
i++
|
||||
}
|
||||
h.T, h.V = h.T[i:], h.V[i:]
|
||||
}
|
||||
}
|
||||
for id := range p.hist { // a container that is gone takes its history with it
|
||||
if strings.HasPrefix(id, prefixOf[name]) && !live[id] && prefixOf[name] != "" {
|
||||
delete(p.hist, id)
|
||||
delete(p.acc, id)
|
||||
}
|
||||
}
|
||||
p.mu.Unlock()
|
||||
|
||||
Reference in New Issue
Block a user