Merge branch 'issues/210' into link-rule

docs/DEV.md

@@ -313,20 +313,54 @@ Kelemetrix is a separate component that consumes audit logs into metric points.
 Tag providers provide string tags to use in kelemetrix.toml,
 and quantifiers provide scalar quantities.
 
-## Testing
+## End-to-end Testing
 
 Kelemetry runs end-to-end tests on GitHub CI.
 Each test suite is a directory under the [`e2e`](../e2e) directory
 with the following structure:
 
-- `client.sh`: Executes a series of operations on a running cluster at `$KUBECONFIG`
-  to trigger generation of a test trace.
-  It can be assumed that the default `make quickstart` configuration is running for the cluster.
-- `config.sh`: Defines the following `local` Bash variables.
+- `config.sh`: Sourced by the e2e test script,
+  executes a series of operations on a running cluster at `$KUBECONFIG`
+  to trigger a background `make quickstart` to generate the test trace.
+  Also defines the following `local` Bash variables.
   It is directly `source`d by the e2e test script.
-  - `TRACE_DISPLAY_NAME`, the display name of the trace shown on the [demo index](https://kubewharf.io/kelemetry/)
+  - `TRACE_DISPLAY_NAME`, the display name of the trace shown on the [demo index](https://kubewharf.io/kelemetry/).
   - `TRACE_SEARCH_TAGS`, an associative array for the search tags to uniquely identify the test trace generated by `client.sh`.
+  - `TEST_DISPLAY_MODE`, the display mode of the trace to test with.
+  - `EXTRA_DISPLAY_MODES` (optional), an indexed array for additional display modes to show on GitHub Pages without getting tested in the e2e test.
 - `validate.jq`: A `jq` script that takes the test trace JSON as input
   and [fails](https://jqlang.github.io/jq/manual/#error) if errors are encountered.
   The script may use library functions defined under [`e2e/lib`](../e2e/lib).
   The current framework uses [gojq](https://github.com/itchyny/gojq/) as the `jq` implementation.
+
+End-to-end tests can be executed locally with the commands:
+
+```console
+$ make kind
+$ make e2e
+```
+
+`make e2e` automatically runs `make quickstart`
+using a custom image built from the local `make output/kelemetry` output.
+If you prefer to run the kelemetry process directly (e.g. with a debugger),
+use the following steps instead:
+
+1. Clean up all previous data by removing the stack volumes completely:
+
+```bash
+make stack COMPOSE_COMMAND='down --remove-orphans --volumes'
+```
+
+2. Recreate the stack:
+
+```bash
+make stack COMPOSE_COMMAND='up --build -d --remove-orphans'
+```
+
+3. Run Kelemetry with your preferred method equivalent to `make`.
+
+4. Execute the e2e test script:
+
+```bash
+bash e2e/run-all.sh
+```

e2e/ancestors/config.sh

@@ -0,0 +1,17 @@
+# script
+kubectl create deployment ancestors --image=alpine:3.16 --replicas=2 -- sleep infinity
+local first_rs_name=$(kubectl get replicasets --selector app=ancestors -o json | jq -r '.items[0].metadata.name')
+sleep 5
+kubectl set image deployments ancestors alpine=alpine:3.17
+sleep 30
+kubectl delete deployments ancestors
+
+# config
+local TEST_DISPLAY_MODE="22000000" # tracing, ancestors only
+local TRACE_DISPLAY_NAME="Deployment"
+
+TRACE_SEARCH_TAGS[cluster]=tracetest
+TRACE_SEARCH_TAGS[group]=apps
+TRACE_SEARCH_TAGS[resource]=replicasets
+TRACE_SEARCH_TAGS[namespace]=default
+TRACE_SEARCH_TAGS[name]=$first_rs_name

e2e/ancestors/validate.jq

@@ -0,0 +1,18 @@
+include "assert";
+include "graph";
+
+.data[0]
+| ._ = (
+  # check root
+  .spans
+    | root
+    | .tags | from_entries
+    | assertEq("root span is deployment"; .resource; "deployments")
+    | assertEq("root span has the correct name"; .name; "ancestors")
+) | ._ = (
+  # root only has one child
+  .spans
+    | children(root.spanID)
+    | assertEq("there is only one child replicaset"; length; 1)
+)
+| null

e2e/deployment/config.sh

@@ -1,6 +1,19 @@
+# script
+kubectl create deployment demo --image=alpine:3.16 --replicas=2 -- sleep infinity
+sleep 5
+kubectl scale deployment demo --replicas=4
+sleep 5
+kubectl set image deployments demo alpine=alpine:3.17
+sleep 5
+kubectl scale deployment demo --replicas=2
+sleep 5
+kubectl delete deployment demo
+sleep 30
+
+# config
+local TEST_DISPLAY_MODE="21000000" # tracing, full tree
 local TRACE_DISPLAY_NAME="Deployment"
 
-declare -A TRACE_SEARCH_TAGS
 TRACE_SEARCH_TAGS[cluster]=tracetest
 TRACE_SEARCH_TAGS[group]=apps
 TRACE_SEARCH_TAGS[resource]=deployments

e2e/run-all.sh

@@ -6,20 +6,16 @@ cd $(dirname $0)
 
 export REPO_PATH=$(realpath ..)
 
-export TEST_DISPLAY_MODE="21000000"
-
-export DISPLAY_MODES="${TEST_DISPLAY_MODE} ${DISPLAY_MODES:-}"
-
 run_test() {
 	local test_name=$1
 
 	local tmpdir=$(mktemp -d)
 
-	if [[ ! -v IS_RERUN ]]; then
-		bash -x ${test_name}/client.sh
-	fi
-
+	declare -A TRACE_SEARCH_TAGS
+	declare -a EXTRA_DISPLAY_MODES
+	set -x
 	source ${test_name}/config.sh
+	set +x
 
 	local curl_param_string=""
 	for curl_param_key in "${!TRACE_SEARCH_TAGS[@]}"; do
@@ -32,6 +28,12 @@ run_test() {
 		-o ${tmpdir}/curl-output.http \
 		localhost:8080/redirect
 
+	if ! (head -n1 ${tmpdir}/curl-output.http | grep '302 Found'); then
+		echo "Trace not found for the parameters"
+		cat curl-output.http
+		exit 1
+	fi
+
 	local full_trace_id=$(grep -P "^Location: /trace/" ${tmpdir}/curl-output.http | cut -d/ -f3 | tr -d '\r')
 	if [[ -z $full_trace_id ]]; then
 		echo "Trace not found for the parameters"
@@ -46,6 +48,8 @@ run_test() {
 		echo "${TRACE_DISPLAY_NAME}" >${OUTPUT_TRACE}/trace-${test_name}/trace_display_name
 		echo ${fixed_id} >${OUTPUT_TRACE}/trace-${test_name}/trace_id
 
+		DISPLAY_MODES=($TEST_DISPLAY_MODE ${EXTRA_DISPLAY_MODES[@]})
+
 		for mode in ${DISPLAY_MODES}; do
 			local mode_trace_id=ff${mode}${fixed_id}
 			curl -o ${OUTPUT_TRACE}/api/traces/${mode_trace_id} localhost:16686/api/traces/${mode_trace_id}
@@ -59,10 +63,12 @@ run_test() {
 
 declare -A pids
 
-for client in */client.sh; do
-	test_name=$(basename $(dirname $client))
-	run_test ${test_name} &
-	pids[$test_name]=$!
+for sh in */validate.jq; do
+	test_name=$(basename $(dirname $sh))
+	if [[ ! -v FILTER_TESTS ]] || [[ $FILTER_TESTS == *$test_name* ]]; then
+		run_test ${test_name} &
+		pids[$test_name]=$!
+	fi
 done
 
 failed_tests=()

e2e/secret-link/config.sh

@@ -1,7 +1,45 @@
+kubectl create -f ${REPO_PATH}/crds/samples/pod-secret.yaml
+
+kubectl create secret generic content
+
+kubectl create -f - << EOF
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: reader
+  labels:
+    app: reader
+spec:
+  selector:
+    matchLabels:
+      app: reader
+  replicas: 2
+  template:
+    metadata:
+      labels:
+        app: reader
+    spec:
+      containers:
+        - name: reader
+          command: ["sleep", "infinity"]
+          image: alpine:3.16
+          volumeMounts:
+            - name: content
+              mountPath: /mnt/content
+      volumes:
+        - name: content
+          secret:
+            secretName: content
+EOF
+sleep 10
+
+kubectl delete deployment/reader secret/content
+sleep 20
+
+# config
+local TEST_DISPLAY_MODE="24100000" # tracing, owned objects, volumes
 local TRACE_DISPLAY_NAME="Volume links"
-local TEST_DISPLAY_MODE="24100000"
 
-declare -A TRACE_SEARCH_TAGS
 TRACE_SEARCH_TAGS[cluster]=tracetest
 TRACE_SEARCH_TAGS[group]=apps
 TRACE_SEARCH_TAGS[resource]=deployments

hack/kind-cluster.yaml

@@ -14,6 +14,7 @@ nodes:
         apiServer:
           extraArgs:
             audit-webhook-config-file: /mnt/audit/audit-kubeconfig.local.yaml
+            audit-webhook-batch-max-wait: 5s # speed up event consumption during test
             tracing-config-file: /mnt/audit/tracing-config.local.yaml
             audit-policy-file: /mnt/audit/audit-policy.yaml
           extraVolumes:

hack/tfconfig.yaml

@@ -63,7 +63,10 @@ modifiers:
     modifierName: link-selector
     args:
       modifierClass: owner-ref
-      includeSiblings: true
+      includeSiblings: false
+      ifAll:
+        - linkClass: children
+          fromParent: false
       upwardDistance: 3
   "04000000":
     # the entire subtree under this object
@@ -77,7 +80,7 @@ modifiers:
       downwardDistance: 3
   "00100000":
     # include secrets under pods
-    displayName: volume secrets
+    displayName: volumes
     modifierName: link-selector
     args:
       ifAll:

pkg/frontend/backend/jaeger-storage/backend.go

@@ -33,6 +33,7 @@ import (
 	"go.uber.org/zap"
 
 	jaegerbackend "github.com/kubewharf/kelemetry/pkg/frontend/backend"
+	jaegerreader "github.com/kubewharf/kelemetry/pkg/frontend/reader"
 	tftree "github.com/kubewharf/kelemetry/pkg/frontend/tf/tree"
 	"github.com/kubewharf/kelemetry/pkg/manager"
 	utiljaeger "github.com/kubewharf/kelemetry/pkg/util/jaeger"
@@ -152,7 +153,15 @@ func (backend *Backend) List(
 	params *spanstore.TraceQueryParameters,
 ) ([]*jaegerbackend.TraceThumbnail, error) {
 	traceThumbnails := []*jaegerbackend.TraceThumbnail{}
-	for _, traceSource := range zconstants.KnownTraceSources(false) {
+
+	var traceSources []string
+	if ctx.Value(jaegerreader.WantPseudoSpansOnly{}) != nil {
+		traceSources = zconstants.KnownPseudoTraceSources()
+	} else {
+		traceSources = zconstants.KnownNonPseudoTraceSources()
+	}
+
+	for _, traceSource := range traceSources {
 		if params.NumTraces != 0 && len(traceThumbnails) >= params.NumTraces {
 			break
 		}

pkg/frontend/reader/merge/merge.go

@@ -614,6 +614,11 @@ func createVirtualSpan(traceId model.TraceID, span *model.Span, opName string, s
 		StartTime:     span.StartTime,
 		Duration:      span.Duration,
 		Tags: []model.KeyValue{
+			{
+				Key:   zconstants.TraceSource,
+				VType: model.StringType,
+				VStr:  zconstants.TraceSourceObject,
+			},
 			{
 				Key:   zconstants.PseudoType,
 				VType: model.StringType,