Skip to content

Made workers' requests channel non-blocking #239

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Oct 29, 2025
Merged

Made workers' requests channel non-blocking #239

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions pkg/llm-d-inference-sim/simulator.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -345,7 +345,7 @@ func (s *VllmSimulator) initializeSim(ctx context.Context) error {
ctx: ctx,
logger: s.logger,
finishedChan: s.workerFinished,
reqChan: make(chan *openaiserverapi.CompletionReqCtx),
reqChan: make(chan *openaiserverapi.CompletionReqCtx, 1),
processor: s,
}
go worker.waitForRequests()
Expand Down Expand Up @@ -402,8 +402,8 @@ func (s *VllmSimulator) processing(ctx context.Context) {
s.logger.Info("Request processing done")
return
case completedReq := <-s.workerFinished:
s.logger.V(4).Info("Worker finished")
worker := completedReq.worker
s.logger.V(4).Info("Worker finished", "worker", worker.id)
s.decrementLora(completedReq.model)
// there is a free worker - find a request for it and send this request for
// processing with this worker
Expand Down
35 changes: 16 additions & 19 deletions pkg/llm-d-inference-sim/worker_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -344,10 +344,15 @@ var _ = Describe("Simulator requests scheduling", Ordered, func() {
option.WithBaseURL(baseURL),
option.WithHTTPClient(client))

var wg sync.WaitGroup

// Run 2000 requests for 2 loras simultaneously
numberOfRequests := 2000
wg.Add(numberOfRequests)

for i := range numberOfRequests {
go func() {
defer wg.Done()
defer GinkgoRecover()
params := openai.ChatCompletionNewParams{
Messages: []openai.ChatCompletionMessageParamUnion{
Expand Down Expand Up @@ -385,8 +390,11 @@ var _ = Describe("Simulator requests scheduling", Ordered, func() {

// After about 2 secs (the mean ttft), send 500 more requests
numberOfRequests = 500
wg.Add(numberOfRequests)

for i := range numberOfRequests {
go func() {
defer wg.Done()
defer GinkgoRecover()
params := openai.ChatCompletionNewParams{
Messages: []openai.ChatCompletionMessageParamUnion{
Expand All @@ -408,15 +416,8 @@ var _ = Describe("Simulator requests scheduling", Ordered, func() {
metrics = strings.Split(string(data), "\n")

// We sent 2500 requests, after about 2.5 seconds
// number of running requests should be 1000
// number of running requests should be about 1000
// and the number of waiting requests should be less than 1000.
// Since we are in the middle of requests scheduling,
// the number of running requests can be 999.
runningStr = findMetric(metrics, runningMetric)
Expect(runningStr).NotTo(Equal(""))
running, err = strconv.Atoi(runningStr)
Expect(err).NotTo(HaveOccurred())
Expect(running).To(Or(Equal(1000), Equal(999)))
waitingStr = findMetric(metrics, waitingMetric)
waiting, err = strconv.Atoi(waitingStr)
Expect(err).NotTo(HaveOccurred())
Expand All @@ -431,19 +432,15 @@ var _ = Describe("Simulator requests scheduling", Ordered, func() {
Expect(err).NotTo(HaveOccurred())
metrics = strings.Split(string(data), "\n")

// The number of running requests should be 1000
// and the number of waiting requests should be less than 1000.
// Since we are in the middle of requests scheduling,
// the number of running requests can be 999.
runningStr = findMetric(metrics, runningMetric)
Expect(runningStr).NotTo(Equal(""))
running, err = strconv.Atoi(runningStr)
Expect(err).NotTo(HaveOccurred())
Expect(running).To(Or(Equal(1000), Equal(999)))
// The number of running requests should be about 1000
// and the number of waiting requests should be less than the
// previous number of waiting requests.
waitingStr = findMetric(metrics, waitingMetric)
waiting, err = strconv.Atoi(waitingStr)
newWaiting, err := strconv.Atoi(waitingStr)
Expect(err).NotTo(HaveOccurred())
Expect(waiting).To(BeNumerically("<", 1000))
Expect(newWaiting).To(BeNumerically("<=", waiting))

wg.Wait()
})
})
})
Expand Down