Files
llmgw/internal/accounting/shared.go
T
Ilker ManapandClaude Opus 5.5 540b518fde
CI / test (push) Successful in 9m30s
Rate limits counted across the control planes with router.active_active
Each router's per-counter request and token counts, this minute's and the
last, ride the standby's two-second exchange with the leader. The leader
answers each standby with everyone else's sum, its own included, and puts
the standbys' sum in force on its own router; each limiter adds the
others' counts to its sliding window.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-11-01 17:30:08 +01:00

111 lines
3.1 KiB
Go

package accounting
import "time"
// Counts are one limiter's counts by counter: this minute's and the one
// before, as absolute figures, so reporting them twice changes nothing.
type Counts struct {
Minute time.Time `json:"minute"`
Current map[string]int64 `json:"current,omitempty"`
Previous map[string]int64 `json:"previous,omitempty"`
}
// Add sums other into c, aligned to c's minute: counts of a minute that is
// c's previous one land there, older ones are dropped. Clocks a second apart
// at a minute's turn are why.
func (c *Counts) Add(other Counts) {
switch {
case other.Minute.Equal(c.Minute):
c.Current = addInto(c.Current, other.Current)
c.Previous = addInto(c.Previous, other.Previous)
case other.Minute.Add(windowLen).Equal(c.Minute):
c.Previous = addInto(c.Previous, other.Current)
}
}
func addInto(dst, src map[string]int64) map[string]int64 {
if len(src) == 0 {
return dst
}
if dst == nil {
dst = make(map[string]int64, len(src))
}
for k, v := range src {
dst[k] += v
}
return dst
}
// Counts reads this limiter's own counts, without the others'.
func (l *Limiter) Counts() Counts {
now := l.now()
out := Counts{Minute: now.Truncate(windowLen), Current: map[string]int64{}, Previous: map[string]int64{}}
l.mu.Lock()
defer l.mu.Unlock()
for k, w := range l.keys {
w.roll(now)
if w.current > 0 {
out.Current[k] = w.current
}
if w.previous > 0 {
out.Previous[k] = w.previous
}
}
return out
}
// SetOthers replaces what the other control planes have counted. A counter
// they did not mention has nothing from them.
func (l *Limiter) SetOthers(c Counts) {
now := l.now()
aligned := Counts{Minute: now.Truncate(windowLen)}
aligned.Add(c)
l.mu.Lock()
defer l.mu.Unlock()
for k, w := range l.keys {
w.roll(now)
w.othersCurrent, w.othersPrevious = aligned.Current[k], aligned.Previous[k]
delete(aligned.Current, k)
delete(aligned.Previous, k)
}
// Counters only others have used yet: a key spread over the cluster is
// limited here from its first request.
for k := range aligned.Current {
w := l.windowLocked(k, now)
w.othersCurrent, w.othersPrevious = aligned.Current[k], aligned.Previous[k]
delete(aligned.Previous, k)
}
for k, v := range aligned.Previous {
l.windowLocked(k, now).othersPrevious = v
}
}
// Rates are an accountant's two limiters' counts.
type Rates struct {
Requests Counts `json:"requests"`
Tokens Counts `json:"tokens"`
}
// Add sums other into r.
func (r *Rates) Add(other Rates) {
r.Requests.Add(other.Requests)
r.Tokens.Add(other.Tokens)
}
// NewRates is an empty sum for the minute now is in.
func NewRates(now time.Time) Rates {
m := now.Truncate(windowLen)
return Rates{Requests: Counts{Minute: m}, Tokens: Counts{Minute: m}}
}
// Rates reads this process's own request and token counts.
func (a *Accountant) Rates() Rates {
return Rates{Requests: a.requests.Counts(), Tokens: a.tokens.Counts()}
}
// SetOtherRates puts the other control planes' counts in force here.
func (a *Accountant) SetOtherRates(r Rates) {
a.requests.SetOthers(r.Requests)
a.tokens.SetOthers(r.Tokens)
}