Skip to content

resource based throttling: reject only adhoc queries #6947

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

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
+107 −64
Open

resource based throttling: reject only adhoc queries #6947

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8bb3d4f
resource based throttling: reject only adhoc queries
erlan-z Aug 8, 2025
7cfa79a
replace request source implementation on qfe when ruler calls qfe to …
erlan-z Aug 15, 2025
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
1 change: 1 addition & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,6 +69,7 @@
* [ENHANCEMENT] Query Frontend: Add a `format_query` label value to the `op` label at `cortex_query_frontend_queries_total` metric. #6925
* [ENHANCEMENT] API: add request ID injection to context to enable tracking requests across downstream services. #6895
* [ENHANCEMENT] gRPC: Add gRPC Channelz monitoring. #6950
* [ENHANCEMENT] Add source metadata to requests(api vs ruler) and used in resource based throttling to only reject adhoc queries. #6947
* [BUGFIX] Ingester: Avoid error or early throttling when READONLY ingesters are present in the ring #6517
* [BUGFIX] Ingester: Fix labelset data race condition. #6573
* [BUGFIX] Compactor: Cleaner should not put deletion marker for blocks with no-compact marker. #6576
Expand Down
1 change: 1 addition & 0 deletions pkg/api/middlewares.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,7 @@ func (h HTTPHeaderMiddleware) injectRequestContext(r *http.Request) *http.Reques
reqId = uuid.NewString()
}
requestContextMap[requestmeta.RequestIdKey] = reqId
requestContextMap[requestmeta.RequestSourceKey] = requestmeta.SourceAPI

ctx := requestmeta.ContextWithRequestMetadataMap(r.Context(), requestContextMap)
return r.WithContext(ctx)
Expand Down
9 changes: 4 additions & 5 deletions pkg/frontend/transport/handler.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@ import (
util_api "github.com/cortexproject/cortex/pkg/util/api"
"github.com/cortexproject/cortex/pkg/util/limiter"
util_log "github.com/cortexproject/cortex/pkg/util/log"
"github.com/cortexproject/cortex/pkg/util/requestmeta"
)

const (
Expand Down Expand Up @@ -245,11 +246,11 @@ func (f *Handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
}

userID := tenant.JoinTenantIDs(tenantIDs)
source := tripperware.GetSource(r)

if f.tenantFederationCfg.Enabled {
maxTenant := f.tenantFederationCfg.MaxTenant
if maxTenant > 0 && len(tenantIDs) > maxTenant {
source := tripperware.GetSource(r.Header.Get("User-Agent"))
if f.cfg.QueryStatsEnabled {
f.rejectedQueries.WithLabelValues(reasonTooManyTenants, source, userID).Inc()
}
Expand Down Expand Up @@ -289,15 +290,13 @@ func (f *Handler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
}
http.Error(w, err.Error(), statusCode)
if f.cfg.QueryStatsEnabled && util.IsRequestBodyTooLarge(err) {
source := tripperware.GetSource(r.Header.Get("User-Agent"))
f.rejectedQueries.WithLabelValues(reasonRequestBodySizeExceeded, source, userID).Inc()
}
return
}
r.Body = io.NopCloser(&buf)
}

source := tripperware.GetSource(r.Header.Get("User-Agent"))
// Log request
if f.cfg.QueryStatsEnabled {
queryString = f.parseRequestQueryString(r, buf)
Expand Down Expand Up @@ -401,7 +400,7 @@ func (f *Handler) logQueryRequest(r *http.Request, queryString url.Values, sourc
logMessage = append(logMessage, "accept_encoding", acceptEncoding)
}

shouldLog := source == tripperware.SourceAPI || (f.cfg.EnabledRulerQueryStatsLog && source == tripperware.SourceRuler)
shouldLog := source == requestmeta.SourceAPI || (f.cfg.EnabledRulerQueryStatsLog && source == requestmeta.SourceRuler)
if shouldLog {
logMessage = append(logMessage, formatQueryString(queryString)...)
level.Info(util_log.WithContext(r.Context(), f.log)).Log(logMessage...)
Expand Down Expand Up @@ -531,7 +530,7 @@ func (f *Handler) reportQueryStats(r *http.Request, source, userID string, query
}
}

shouldLog := source == tripperware.SourceAPI || (f.cfg.EnabledRulerQueryStatsLog && source == tripperware.SourceRuler)
shouldLog := source == requestmeta.SourceAPI || (f.cfg.EnabledRulerQueryStatsLog && source == requestmeta.SourceRuler)
if shouldLog {
logMessage = append(logMessage, formatQueryString(queryString)...)
if error != nil {
Expand Down
51 changes: 26 additions & 25 deletions pkg/frontend/transport/handler_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,11 +26,11 @@ import (

querier_stats "github.com/cortexproject/cortex/pkg/querier/stats"
"github.com/cortexproject/cortex/pkg/querier/tenantfederation"
"github.com/cortexproject/cortex/pkg/querier/tripperware"
"github.com/cortexproject/cortex/pkg/tenant"
util_api "github.com/cortexproject/cortex/pkg/util/api"
"github.com/cortexproject/cortex/pkg/util/limiter"
util_log "github.com/cortexproject/cortex/pkg/util/log"
"github.com/cortexproject/cortex/pkg/util/requestmeta"
)

type roundTripperFunc func(*http.Request) (*http.Response, error)
Expand Down Expand Up @@ -216,7 +216,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonResponseBodySizeExceeded, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonResponseBodySizeExceeded, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusRequestEntityTooLarge,
Expand All @@ -232,7 +232,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonTooManyRequests, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonTooManyRequests, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusTooManyRequests,
Expand All @@ -248,7 +248,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonTooManySamples, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonTooManySamples, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -264,7 +264,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonTimeRangeExceeded, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonTimeRangeExceeded, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -280,7 +280,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonSeriesFetched, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonSeriesFetched, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -296,7 +296,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonChunksFetched, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonChunksFetched, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -312,7 +312,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonChunkBytesFetched, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonChunkBytesFetched, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -328,7 +328,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonDataBytesFetched, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonDataBytesFetched, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -344,7 +344,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonSeriesLimitStoreGateway, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonSeriesLimitStoreGateway, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -360,7 +360,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonChunksLimitStoreGateway, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonChunksLimitStoreGateway, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -376,7 +376,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonBytesLimitStoreGateway, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonBytesLimitStoreGateway, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -393,7 +393,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonResourceExhausted, tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.rejectedQueries.WithLabelValues(reasonResourceExhausted, requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusUnprocessableEntity,
Expand All @@ -410,7 +410,7 @@ func TestHandler_ServeHTTP(t *testing.T) {
}, nil
}),
additionalMetricsCheckFunc: func(h *Handler) {
v := promtest.ToFloat64(h.slowQueries.WithLabelValues(tripperware.SourceAPI, userID))
v := promtest.ToFloat64(h.slowQueries.WithLabelValues(requestmeta.SourceAPI, userID))
assert.Equal(t, float64(1), v)
},
expectedStatusCode: http.StatusOK,
Expand Down Expand Up @@ -472,12 +472,12 @@ func TestReportQueryStatsFormat(t *testing.T) {
tests := map[string]testCase{
"should not include query and header details if empty": {
expectedLog: `level=info msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=0 response_series_count=0 fetched_series_count=0 fetched_chunks_count=0 fetched_samples_count=0 fetched_chunks_bytes=0 fetched_data_bytes=0 split_queries=0 status_code=200 response_size=1000 samples_scanned=0`,
source: tripperware.SourceAPI,
source: requestmeta.SourceAPI,
},
"should include query length and string at the end": {
queryString: url.Values(map[string][]string{"query": {"up"}}),
expectedLog: `level=info msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=0 response_series_count=0 fetched_series_count=0 fetched_chunks_count=0 fetched_samples_count=0 fetched_chunks_bytes=0 fetched_data_bytes=0 split_queries=0 status_code=200 response_size=1000 samples_scanned=0 query_length=2 param_query=up`,
source: tripperware.SourceAPI,
source: requestmeta.SourceAPI,
},
"should include query stats": {
queryStats: &querier_stats.QueryStats{
Expand All @@ -494,17 +494,17 @@ func TestReportQueryStatsFormat(t *testing.T) {
},
},
expectedLog: `level=info msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=3 response_series_count=100 fetched_series_count=100 fetched_chunks_count=200 fetched_samples_count=300 fetched_chunks_bytes=1024 fetched_data_bytes=2048 split_queries=10 status_code=200 response_size=1000 samples_scanned=0 query_storage_wall_time_seconds=6000`,
source: tripperware.SourceAPI,
source: requestmeta.SourceAPI,
},
"should include user agent": {
header: http.Header{"User-Agent": []string{"Grafana"}},
expectedLog: `level=info msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=0 response_series_count=0 fetched_series_count=0 fetched_chunks_count=0 fetched_samples_count=0 fetched_chunks_bytes=0 fetched_data_bytes=0 split_queries=0 status_code=200 response_size=1000 samples_scanned=0 user_agent=Grafana`,
source: tripperware.SourceAPI,
source: requestmeta.SourceAPI,
},
"should include response error": {
responseErr: errors.New("foo_err"),
expectedLog: `level=error msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=0 response_series_count=0 fetched_series_count=0 fetched_chunks_count=0 fetched_samples_count=0 fetched_chunks_bytes=0 fetched_data_bytes=0 split_queries=0 status_code=200 response_size=1000 samples_scanned=0 error=foo_err`,
source: tripperware.SourceAPI,
source: requestmeta.SourceAPI,
},
"should include query priority": {
queryString: url.Values(map[string][]string{"query": {"up"}}),
Expand All @@ -513,7 +513,7 @@ func TestReportQueryStatsFormat(t *testing.T) {
PriorityAssigned: true,
},
expectedLog: `level=info msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=0 response_series_count=0 fetched_series_count=0 fetched_chunks_count=0 fetched_samples_count=0 fetched_chunks_bytes=0 fetched_data_bytes=0 split_queries=0 status_code=200 response_size=1000 samples_scanned=0 query_length=2 priority=99 param_query=up`,
source: tripperware.SourceAPI,
source: requestmeta.SourceAPI,
},
"should include data fetch min and max time": {
queryString: url.Values(map[string][]string{"query": {"up"}}),
Expand All @@ -522,7 +522,7 @@ func TestReportQueryStatsFormat(t *testing.T) {
DataSelectMinTime: 1704067200000,
},
expectedLog: `level=info msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=0 response_series_count=0 fetched_series_count=0 fetched_chunks_count=0 fetched_samples_count=0 fetched_chunks_bytes=0 fetched_data_bytes=0 split_queries=0 status_code=200 response_size=1000 samples_scanned=0 data_select_max_time=1704153600 data_select_min_time=1704067200 query_length=2 param_query=up`,
source: tripperware.SourceAPI,
source: requestmeta.SourceAPI,
},
"should include query stats with store gateway stats": {
queryStats: &querier_stats.QueryStats{
Expand All @@ -541,16 +541,16 @@ func TestReportQueryStatsFormat(t *testing.T) {
},
},
expectedLog: `level=info msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=3 response_series_count=100 fetched_series_count=100 fetched_chunks_count=200 fetched_samples_count=300 fetched_chunks_bytes=1024 fetched_data_bytes=2048 split_queries=10 status_code=200 response_size=1000 samples_scanned=0 store_gateway_touched_postings_count=20 store_gateway_touched_posting_bytes=200 query_storage_wall_time_seconds=6000`,
source: tripperware.SourceAPI,
source: requestmeta.SourceAPI,
},
"should not report a log": {
expectedLog: ``,
source: tripperware.SourceRuler,
source: requestmeta.SourceRuler,
enabledRulerQueryStatsLog: false,
},
"should report a log": {
expectedLog: `level=info msg="query stats" component=query-frontend method=GET path=/prometheus/api/v1/query response_time=1s query_wall_time_seconds=0 response_series_count=0 fetched_series_count=0 fetched_chunks_count=0 fetched_samples_count=0 fetched_chunks_bytes=0 fetched_data_bytes=0 split_queries=0 status_code=200 response_size=1000 samples_scanned=0`,
source: tripperware.SourceRuler,
source: requestmeta.SourceRuler,
enabledRulerQueryStatsLog: true,
},
}
Expand All @@ -559,6 +559,7 @@ func TestReportQueryStatsFormat(t *testing.T) {
t.Run(testName, func(t *testing.T) {
handler := NewHandler(HandlerConfig{QueryStatsEnabled: true, EnabledRulerQueryStatsLog: testData.enabledRulerQueryStatsLog}, tenantfederation.Config{}, http.DefaultTransport, logger, nil)
req.Header = testData.header
req = req.WithContext(requestmeta.ContextWithRequestSource(context.Background(), testData.source))
handler.reportQueryStats(req, testData.source, userID, testData.queryString, responseTime, testData.queryStats, testData.responseErr, statusCode, resp)
data, err := io.ReadAll(outputBuf)
require.NoError(t, err)
Expand Down Expand Up @@ -706,7 +707,7 @@ func Test_TenantFederation_MaxTenant(t *testing.T) {
require.Contains(t, string(body), test.expectedErrMsg)

if strings.Contains(test.expectedErrMsg, "too many tenants") {
v := promtest.ToFloat64(handler.rejectedQueries.WithLabelValues(reasonTooManyTenants, tripperware.SourceAPI, test.orgId))
v := promtest.ToFloat64(handler.rejectedQueries.WithLabelValues(reasonTooManyTenants, requestmeta.SourceAPI, test.orgId))
assert.Equal(t, float64(1), v)
}
}
Expand Down
13 changes: 7 additions & 6 deletions pkg/ingester/ingester.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,7 @@ import (
"github.com/cortexproject/cortex/pkg/util/limiter"
logutil "github.com/cortexproject/cortex/pkg/util/log"
util_math "github.com/cortexproject/cortex/pkg/util/math"
"github.com/cortexproject/cortex/pkg/util/requestmeta"
"github.com/cortexproject/cortex/pkg/util/resource"
"github.com/cortexproject/cortex/pkg/util/services"
"github.com/cortexproject/cortex/pkg/util/spanlogger"
Expand Down Expand Up @@ -1696,7 +1697,7 @@ func (i *Ingester) QueryExemplars(ctx context.Context, req *client.ExemplarQuery
}

// We will report *this* request in the error too.
c, err := i.trackInflightQueryRequest()
c, err := i.trackInflightQueryRequest(ctx)
if err != nil {
return nil, err
}
Expand Down Expand Up @@ -1804,7 +1805,7 @@ func (i *Ingester) labelsValuesCommon(ctx context.Context, req *client.LabelValu
q.Close()
}

c, err := i.trackInflightQueryRequest()
c, err := i.trackInflightQueryRequest(ctx)
if err != nil {
return nil, cleanup, err
}
Expand Down Expand Up @@ -1901,7 +1902,7 @@ func (i *Ingester) labelNamesCommon(ctx context.Context, req *client.LabelNamesR
q.Close()
}

c, err := i.trackInflightQueryRequest()
c, err := i.trackInflightQueryRequest(ctx)
if err != nil {
return nil, cleanup, err
}
Expand Down Expand Up @@ -2252,7 +2253,7 @@ func (i *Ingester) QueryStream(req *client.QueryRequest, stream client.Ingester_
return nil
}

func (i *Ingester) trackInflightQueryRequest() (func(), error) {
func (i *Ingester) trackInflightQueryRequest(ctx context.Context) (func(), error) {
gl := i.getInstanceLimits()
if gl != nil && gl.MaxInflightQueryRequests > 0 {
if i.inflightQueryRequests.Load() >= gl.MaxInflightQueryRequests {
Expand All @@ -2262,7 +2263,7 @@ func (i *Ingester) trackInflightQueryRequest() (func(), error) {

i.maxInflightQueryRequests.Track(i.inflightQueryRequests.Inc())

if i.resourceBasedLimiter != nil {
if i.resourceBasedLimiter != nil && !requestmeta.RequestFromRuler(ctx) {
if err := i.resourceBasedLimiter.AcceptNewRequest(); err != nil {
level.Warn(i.logger).Log("msg", "failed to accept request", "err", err)
return nil, httpgrpc.Errorf(http.StatusServiceUnavailable, "failed to query: %s", limiter.ErrResourceLimitReachedStr)
Expand All @@ -2282,7 +2283,7 @@ func (i *Ingester) queryStreamChunks(ctx context.Context, db *userTSDB, from, th
}
defer q.Close()

c, err := i.trackInflightQueryRequest()
c, err := i.trackInflightQueryRequest(ctx)
if err != nil {
return 0, 0, 0, 0, err
}
Expand Down
Loading
Loading