Skip to content

Fixed TestSchedulerEnqueueWithFrontendDisconnect flakyness #3434

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.

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
Nov 2, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
63 changes: 41 additions & 22 deletions pkg/querier/frontend2/scheduler.go
Original file line number Diff line number Diff line change
Expand Up @@ -42,9 +42,10 @@ type Scheduler struct {
pendingRequests map[requestKey]*schedulerRequest // Request is kept in this map even after being dispatched to querier. It can still be canceled at that time.

// Metrics.
connectedWorkers prometheus.GaugeFunc
queueDuration prometheus.Histogram
queueLength *prometheus.GaugeVec
connectedQuerierClients prometheus.GaugeFunc
connectedFrontendClients prometheus.GaugeFunc
queueDuration prometheus.Histogram
queueLength *prometheus.GaugeVec
}

type requestKey struct {
Expand All @@ -71,33 +72,35 @@ func (cfg *SchedulerConfig) RegisterFlags(f *flag.FlagSet) {

// NewScheduler creates a new Scheduler.
func NewScheduler(cfg SchedulerConfig, limits Limits, log log.Logger, registerer prometheus.Registerer) (*Scheduler, error) {
connectedQuerierWorkers := atomic.NewInt32(0)
s := &Scheduler{
log: log,
limits: limits,

queues: newUserQueues(cfg.MaxOutstandingPerTenant),
pendingRequests: map[requestKey]*schedulerRequest{},

queueDuration: promauto.With(registerer).NewHistogram(prometheus.HistogramOpts{
Name: "cortex_query_scheduler_queue_duration_seconds",
Help: "Time spend by requests in queue before getting picked up by a querier.",
Buckets: prometheus.DefBuckets,
}),
connectedWorkers: promauto.With(registerer).NewGaugeFunc(prometheus.GaugeOpts{
Name: "cortex_query_scheduler_connected_workers",
Help: "Number of querier worker clients currently connected to the query-scheduler.",
}, func() float64 { return float64(connectedQuerierWorkers.Load()) }),
queueLength: promauto.With(registerer).NewGaugeVec(prometheus.GaugeOpts{
Name: "cortex_query_scheduler_queue_length",
Help: "Number of queries in the queue.",
}, []string{"user"}),

queues: newUserQueues(cfg.MaxOutstandingPerTenant),
pendingRequests: map[requestKey]*schedulerRequest{},
connectedFrontends: map[string]*connectedFrontend{},
connectedQuerierWorkers: connectedQuerierWorkers,
connectedQuerierWorkers: atomic.NewInt32(0),
}
s.cond = sync.NewCond(&s.mtx)

s.queueDuration = promauto.With(registerer).NewHistogram(prometheus.HistogramOpts{
Name: "cortex_query_scheduler_queue_duration_seconds",
Help: "Time spend by requests in queue before getting picked up by a querier.",
Buckets: prometheus.DefBuckets,
})
s.connectedQuerierClients = promauto.With(registerer).NewGaugeFunc(prometheus.GaugeOpts{
Name: "cortex_query_scheduler_connected_querier_clients",
Help: "Number of querier worker clients currently connected to the query-scheduler.",
}, s.getConnectedQuerierClientsMetric)
s.connectedFrontendClients = promauto.With(registerer).NewGaugeFunc(prometheus.GaugeOpts{
Name: "cortex_query_scheduler_connected_frontend_clients",
Help: "Number of query-frontend worker clients currently connected to the query-scheduler.",
}, s.getConnectedFrontendClientsMetric)
s.queueLength = promauto.With(registerer).NewGaugeVec(prometheus.GaugeOpts{
Name: "cortex_query_scheduler_queue_length",
Help: "Number of queries in the queue.",
}, []string{"user"})

s.Service = services.NewIdleService(nil, s.stopping)
return s, nil
}
Expand Down Expand Up @@ -515,3 +518,19 @@ func (s *Scheduler) unregisterQuerierConnection(querier string) {
defer s.mtx.Unlock()
s.queues.removeQuerierConnection(querier)
}

func (s *Scheduler) getConnectedQuerierClientsMetric() float64 {
return float64(s.connectedQuerierWorkers.Load())
}

func (s *Scheduler) getConnectedFrontendClientsMetric() float64 {
s.connectedFrontendsMu.Lock()
defer s.connectedFrontendsMu.Unlock()

count := 0
for _, workers := range s.connectedFrontends {
count += workers.connections
}

return float64(count)
}
13 changes: 12 additions & 1 deletion pkg/querier/frontend2/scheduler_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@ import (

"github.com/go-kit/kit/log"
"github.com/opentracing/opentracing-go"
promtest "github.com/prometheus/client_golang/prometheus/testutil"
"github.com/stretchr/testify/require"
"github.com/uber/jaeger-client-go/config"
"github.com/weaveworks/common/httpgrpc"
Expand Down Expand Up @@ -164,11 +165,21 @@ func TestSchedulerEnqueueWithFrontendDisconnect(t *testing.T) {
HttpRequest: &httpgrpc.HTTPRequest{Method: "GET", Url: "/hello"},
})

querierLoop := initQuerierLoop(t, querierClient, "querier-1")
// Wait until the frontend has connected to the scheduler.
test.Poll(t, time.Second, float64(1), func() interface{} {
return promtest.ToFloat64(scheduler.connectedFrontendClients)
})

// Disconnect frontend.
require.NoError(t, frontendLoop.CloseSend())

// Wait until the frontend has disconnected.
test.Poll(t, time.Second, float64(0), func() interface{} {
return promtest.ToFloat64(scheduler.connectedFrontendClients)
})

querierLoop := initQuerierLoop(t, querierClient, "querier-1")

verifyQuerierDoesntReceiveRequest(t, querierLoop, 500*time.Millisecond)
verifyNoPendingRequestsLeft(t, scheduler)
}
Expand Down