Compare commits

..

1 Commits

Author SHA1 Message Date
Nikhil Soni
9f03fea0f3 fix(querybuilder): prefer resource over any context for ambiguous filter keys (#12888)
Some checks are pending
build-staging / staging (push) Blocked by required conditions
build-staging / prepare (push) Waiting to run
build-staging / js-build (push) Blocked by required conditions
build-staging / go-build (push) Blocked by required conditions
cacheci / tests (push) Waiting to run
Release Drafter / update_release_draft (push) Waiting to run
#### Description

- A logs filter on a bare key that lives in **both** resource and
another context (body or scope) ANDed the two: the resource candidate
built the `__resource_filter` fingerprint CTE while the other candidate
landed as a required main-query term, so the query matched almost
nothing.
- `ResolveLogicalFields` only preferred resource over `attribute`.
Generalized it to prefer resource over **any** other context (attribute,
body, scope, …); other contexts stay reachable via their qualified names
(e.g. `body.service.name`).

#### Issues closed by this PR

Closes SigNoz/engineering-pod#6086
Part of https://github.com/SigNoz/platform-pod/issues/3158

#### Additional Information

Generalized rather than special-casing body/scope, since any future
context would hit the same fingerprint-CTE trap.
2026-09-17 14:45:45 +00:00
5 changed files with 231 additions and 17 deletions

View File

@@ -25,8 +25,9 @@ const (
// ResolveLogicalFields picks which logical fields a filter term builds conditions
// for. With 0 or 1 field it returns the input unchanged and no warning. When a
// name is ambiguous (several logical fields — a family is one field and never
// ambiguous with itself) it returns a warning; a resource+attribute mix defaults
// to the resource fields (the common intent), noted in the warning.
// ambiguous with itself) it returns a warning; a resource + other-context mix
// (attribute, body, scope, …) defaults to the resource fields (the common
// intent), noted in the warning.
func ResolveLogicalFields(field *telemetrytypes.TelemetryFieldKey, logicalFields []*telemetrytypes.LogicalField) ([]*telemetrytypes.LogicalField, string) {
if len(logicalFields) <= 1 {
return logicalFields, ""
@@ -39,18 +40,17 @@ func ResolveLogicalFields(field *telemetrytypes.TelemetryFieldKey, logicalFields
logicalFields,
)
hasResource, hasAttribute := false, false
hasResource, hasOther := false, false
for _, item := range logicalFields {
switch item.FieldContext {
case telemetrytypes.FieldContextResource:
if item.FieldContext == telemetrytypes.FieldContextResource {
hasResource = true
case telemetrytypes.FieldContextAttribute:
hasAttribute = true
} else {
hasOther = true
}
}
// when there is both resource and attribute context, default to resource only
if hasResource && hasAttribute {
// with resource and any other context, default to resource only
if hasResource && hasOther {
filtered := make([]*telemetrytypes.LogicalField, 0, len(logicalFields))
for _, item := range logicalFields {
if item.FieldContext == telemetrytypes.FieldContextResource {
@@ -58,8 +58,8 @@ func ResolveLogicalFields(field *telemetrytypes.TelemetryFieldKey, logicalFields
}
}
logicalFields = filtered
warning += " " + "Using `resource` context by default. To query attributes explicitly, " +
fmt.Sprintf("use the fully qualified name (e.g., 'attribute.%s')", field.Name)
warning += " " + "Using `resource` context by default. To query another context explicitly, " +
fmt.Sprintf("use the fully qualified name (e.g., 'attribute.%s' or 'body.%s')", field.Name, field.Name)
}
return logicalFields, warning

View File

@@ -175,6 +175,42 @@ func TestResolveLogicalFieldsKeepsFamilyThroughAmbiguity(t *testing.T) {
assert.Equal(t, []string{"deployment.environment.name", "deployment.environment"}, memberNames(resolved[0]))
}
// Resource wins over every other context, not just attribute: a bare key that
// also lives in body or scope must collapse to resource alone, so the surviving
// candidate does not AND against the resource fingerprint CTE.
func TestResolveLogicalFieldsResourceWinsOverOtherContexts(t *testing.T) {
testCases := []struct {
name string
other telemetrytypes.FieldContext
}{
{name: "ResourceOverBody", other: telemetrytypes.FieldContextBody},
{name: "ResourceOverScope", other: telemetrytypes.FieldContextScope},
}
for _, testCase := range testCases {
t.Run(testCase.name, func(t *testing.T) {
requested := &telemetrytypes.TelemetryFieldKey{Name: "service.name"}
fields := []*telemetrytypes.LogicalField{
telemetrytypes.SingleLogicalField("service.name", &telemetrytypes.TelemetryFieldKey{
Name: "service.name",
FieldContext: telemetrytypes.FieldContextResource,
FieldDataType: telemetrytypes.FieldDataTypeString,
}),
telemetrytypes.SingleLogicalField("service.name", &telemetrytypes.TelemetryFieldKey{
Name: "service.name",
FieldContext: testCase.other,
FieldDataType: telemetrytypes.FieldDataTypeString,
}),
}
resolved, warning := ResolveLogicalFields(requested, fields)
assert.NotEmpty(t, warning)
require.Len(t, resolved, 1)
assert.Equal(t, telemetrytypes.FieldContextResource, resolved[0].FieldContext)
})
}
}
// Members of a family with different data types never merge: the identity
// (signal, context, data type) separates them into distinct logical fields.
func TestMatchingLogicalFieldsNeverMergesAcrossDataTypes(t *testing.T) {

View File

@@ -0,0 +1,90 @@
package logsstatementbuilder
import (
"context"
"testing"
"github.com/SigNoz/signoz/pkg/flagger/flaggertest"
"github.com/SigNoz/signoz/pkg/instrumentation/instrumentationtest"
"github.com/SigNoz/signoz/pkg/querybuilder"
"github.com/SigNoz/signoz/pkg/statementbuilder"
"github.com/SigNoz/signoz/pkg/telemetryschema/logstelemetryschema"
qbtypes "github.com/SigNoz/signoz/pkg/types/querybuildertypes/querybuildertypesv5"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes"
"github.com/SigNoz/signoz/pkg/types/telemetrytypes/telemetrytypestest"
"github.com/SigNoz/signoz/pkg/valuer"
"github.com/stretchr/testify/require"
)
// A key present in both resource and body contexts must filter on resource only.
// The resource condition builds the fingerprint CTE, so a surviving body condition
// would AND against it and match almost nothing (engineering-pod#6086).
func TestStatementBuilderResourceBodyConflict(t *testing.T) {
store := telemetrytypestest.NewMockMetadataStore()
store.SetStaticFields(logstelemetryschema.IntrinsicFields)
store.SetKey(&telemetrytypes.TelemetryFieldKey{
Name: "service.name",
Signal: telemetrytypes.SignalLogs,
FieldContext: telemetrytypes.FieldContextResource,
FieldDataType: telemetrytypes.FieldDataTypeString,
})
bodyKey := &telemetrytypes.TelemetryFieldKey{
Name: "service.name",
Signal: telemetrytypes.SignalLogs,
FieldContext: telemetrytypes.FieldContextBody,
FieldDataType: telemetrytypes.FieldDataTypeString,
}
require.NoError(t, bodyKey.SetJSONAccessPlan(telemetrytypes.JSONColumnMetadata{
BaseColumn: logstelemetryschema.LogsV2BodyV2Column,
PromotedColumn: logstelemetryschema.LogsV2BodyPromotedColumn,
}, map[string][]telemetrytypes.FieldDataType{"service.name": {telemetrytypes.FieldDataTypeString}}))
store.SetKey(bodyKey)
fl := flaggertest.WithUseJSONBody(t, true)
storage := logstelemetryschema.NewStorage()
aggExprRewriter := querybuilder.NewAggExprRewriter(instrumentationtest.New().ToProviderSettings(), nil, storage, fl, telemetrytypes.SignalLogs)
statementBuilder := NewLogQueryStatementBuilder(
instrumentationtest.New().ToProviderSettings(),
store,
storage,
aggExprRewriter,
logstelemetryschema.DefaultFullTextColumn,
fl,
nil,
statementbuilder.Config{SkipResourceFingerprint: statementbuilder.SkipResourceFingerprint{Enabled: false, Threshold: 100000}},
)
testCases := []struct {
name string
requestType qbtypes.RequestType
query qbtypes.QueryBuilderQuery[qbtypes.LogAggregation]
expected qbtypes.Statement
}{
{
name: "AmbiguousKeyFiltersResourceOnly",
requestType: qbtypes.RequestTypeRaw,
query: qbtypes.QueryBuilderQuery[qbtypes.LogAggregation]{
Signal: telemetrytypes.SignalLogs,
Filter: &qbtypes.Filter{Expression: "service.name = 'webapp'"},
Limit: 10,
},
expected: qbtypes.Statement{
Query: "WITH __resource_filter AS (SELECT fingerprint FROM signoz_logs.distributed_logs_v2_resource WHERE (simpleJSONExtractString(labels, 'service.name') = ? AND labels LIKE ? AND labels LIKE ?) AND seen_at_ts_bucket_start >= ? AND seen_at_ts_bucket_start <= ? GROUP BY fingerprint) SELECT timestamp, id, trace_id, span_id, trace_flags, severity_text, severity_number, scope_name, scope_version, body_v2 as body, attributes_string, attributes_number, attributes_bool, resources_string, scope_string FROM signoz_logs.distributed_logs_v2 WHERE resource_fingerprint GLOBAL IN (SELECT fingerprint FROM __resource_filter) AND timestamp >= ? AND ts_bucket_start >= ? AND timestamp < ? AND ts_bucket_start <= ? LIMIT ?",
Args: []any{"webapp", "%service.name%", "%service.name\":\"webapp%", uint64(1747945619), uint64(1747983448), "1747947419000000000", uint64(1747945619), "1747983448000000000", uint64(1747983448), 10},
Warnings: []string{
"Key `service.name` is ambiguous, found 2 different combinations of field context / data type: [name=service.name,context=resource,datatype=string name=service.name,context=body,datatype=string]. Using `resource` context by default. To query another context explicitly, use the fully qualified name (e.g., 'attribute.service.name' or 'body.service.name')",
},
},
},
}
for _, testCase := range testCases {
t.Run(testCase.name, func(t *testing.T) {
q, err := statementBuilder.Build(context.Background(), valuer.UUID{}, 1747947419000, 1747983448000, testCase.requestType, testCase.query, nil)
require.NoError(t, err)
require.Equal(t, testCase.expected.Query, q.Query)
require.Equal(t, testCase.expected.Args, q.Args)
require.Equal(t, testCase.expected.Warnings, q.Warnings)
})
}
}

View File

@@ -0,0 +1,88 @@
import json
from collections.abc import Callable
from datetime import UTC, datetime, timedelta
from http import HTTPStatus
from fixtures import types
from fixtures.auth import USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD
from fixtures.logs import Logs
from fixtures.querier import (
build_raw_query,
get_rows,
make_query_request,
)
def test_resource_body_conflict(
signoz: types.SigNoz,
create_user_admin: None, # pylint: disable=unused-argument
get_token: Callable[[str, str], str],
insert_logs: Callable[[list[Logs]], None],
export_json_types: Callable[[list[Logs]], None],
) -> None:
now = datetime.now(tz=UTC)
start_ms = int((now - timedelta(seconds=10)).timestamp() * 1000)
end_ms = int(now.timestamp() * 1000)
# python's body carries service.name, making the bare key ambiguous across
# resource and body; java's body omits it, so ANDing body in would drop it.
logs_list = [
Logs(
timestamp=now - timedelta(seconds=2),
resources={"service.name": "java"},
body_v2=json.dumps({"msg": "hello"}),
body_promoted="",
),
Logs(
timestamp=now - timedelta(seconds=1),
resources={"service.name": "python"},
body_v2=json.dumps({"service.name": "python"}),
body_promoted="",
),
]
export_json_types(logs_list)
insert_logs(logs_list)
token = get_token(USER_ADMIN_EMAIL, USER_ADMIN_PASSWORD)
cases = [
{
"name": "bare_key_resolves_to_resource",
"filter": "service.name = 'java'",
"expected_service_names": ["java"],
"expect_resource_warning": True,
},
{
"name": "qualified_body_key_targets_body",
"filter": "body.service.name = 'python'",
"expected_service_names": ["python"],
"expect_resource_warning": False,
},
]
for case in cases:
response = make_query_request(
signoz,
token,
start_ms,
end_ms,
request_type="raw",
queries=[
build_raw_query(
name="A",
signal="logs",
filter_expression=case["filter"],
limit=100,
step_interval=60,
)
],
)
assert response.status_code == HTTPStatus.OK, f"{case['name']}: {response.text}"
rows = get_rows(response)
assert [row["data"]["resources_string"].get("service.name") for row in rows] == case["expected_service_names"], f"{case['name']}: {response.json()}"
warning = response.json()["data"].get("warning")
if case["expect_resource_warning"]:
assert warning is not None and "Using `resource` context by default" in warning["warnings"][0]["message"], f"{case['name']}: {warning}"
else:
assert warning is None, f"{case['name']}: {warning}"

View File

@@ -64,8 +64,8 @@ def test_resource_default_warning(
"Key `service.name` is ambiguous, found 2 different combinations of "
"field context / data type: [name=service.name,context=resource,datatype=string "
"name=service.name,context=attribute,datatype=string]. Using `resource` context "
"by default. To query attributes explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name')"
"by default. To query another context explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name' or 'body.service.name')"
)
assert warning["warnings"] == [
{"message": expected_service_name_warning},
@@ -237,8 +237,8 @@ def test_deduped_warnings_for_single_query(
"Key `service.name` is ambiguous, found 2 different combinations of "
"field context / data type: [name=service.name,context=resource,datatype=string "
"name=service.name,context=attribute,datatype=string]. Using `resource` context "
"by default. To query attributes explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name')"
"by default. To query another context explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name' or 'body.service.name')"
)
expected_status_code_warning = "Key `http.status_code` is ambiguous, found 2 different combinations of field context / data type: [name=http.status_code,context=attribute,datatype=number name=http.status_code,context=attribute,datatype=string]."
assert warning["warnings"] == [
@@ -328,8 +328,8 @@ def test_deduped_warnings_for_multiple_queries(
"Key `service.name` is ambiguous, found 2 different combinations of "
"field context / data type: [name=service.name,context=resource,datatype=string "
"name=service.name,context=attribute,datatype=string]. Using `resource` context "
"by default. To query attributes explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name')"
"by default. To query another context explicitly, use the fully qualified name "
"(e.g., 'attribute.service.name' or 'body.service.name')"
)
expected_status_code_warning = "Key `http.status_code` is ambiguous, found 2 different combinations of field context / data type: [name=http.status_code,context=attribute,datatype=number name=http.status_code,context=attribute,datatype=string]."
assert warning["warnings"] == [