github/troubleshoot
1
0
Fork 0
mirror of https://github.com/replicatedhq/troubleshoot.git synced 2026-08-27 00:37:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(analyze): warn when nodeResources has no node data, add ignoreIfNoFiles (#2035)

Browse Source 
* feat(analyze): warn when nodeResources has no node data, add ignoreIfNoFiles

The nodeResources analyzer previously failed silently when
cluster-resources/nodes.json was not collected (e.g. when the
clusterResources collector was excluded or could not list nodes).
It now emits a warn outcome per nodeResources entry in the spec.

Add an ignoreIfNoFiles top-level field to nodeResources, mirroring
textAnalyze, so users can opt out of the new warning when the
analyzer is intentionally optional.

- Add IgnoreIfNoFiles to v1beta2.NodeResources
- Update CRDs and JSON schemas
- Unit test the warn / ignore paths
- Add an e2e fixture and test that excludes clusterResources and
asserts the analyze output

Signed-off-by: Evans Mungai <evans@replicated.com>

* Fix review comment

Signed-off-by: Evans Mungai <evans@replicated.com>

---------

Signed-off-by: Evans Mungai <evans@replicated.com>
This commit is contained in:
Evans Mungai
2026-04-29 13:13:24 +01:00
committed by GitHub
parent 572cb8e31d
commit 2fdd21fca1
11 changed files with 229 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
config/crds/troubleshoot.sh_analyzers.yaml
+2
View File
@@ -1360,6 +1360,8 @@ spec:
- key
type: object
type: object
ignoreIfNoFiles:
type: boolean
outcomes:
items:
properties:
config/crds/troubleshoot.sh_preflights.yaml
+2
View File
@@ -1360,6 +1360,8 @@ spec:
- key
type: object
type: object
ignoreIfNoFiles:
type: boolean
outcomes:
items:
properties:
config/crds/troubleshoot.sh_supportbundles.yaml
+2
View File
@@ -1391,6 +1391,8 @@ spec:
- key
type: object
type: object
ignoreIfNoFiles:
type: boolean
outcomes:
items:
properties:
pkg/analyze/node_resources.go
+17 -1
View File
@@ -17,6 +17,7 @@ import (
troubleshootv1beta2 "github.com/replicatedhq/troubleshoot/pkg/apis/troubleshoot/v1beta2"
"github.com/replicatedhq/troubleshoot/pkg/constants"
"github.com/replicatedhq/troubleshoot/pkg/k8sutil"
"github.com/replicatedhq/troubleshoot/pkg/types"
)
type AnalyzeNodeResources struct {
@@ -45,6 +46,9 @@ func (a *AnalyzeNodeResources) Analyze(getFile getCollectedFileContents, findFil
if err != nil {
return nil, err
}
if result == nil {
return nil, nil
}
result.Strict = a.analyzer.Strict.BoolOrDefaultFalse()
return []*AnalyzeResult{result}, nil
}
@@ -53,7 +57,19 @@ func (a *AnalyzeNodeResources) analyzeNodeResources(analyzer *troubleshootv1beta
collected, err := getCollectedFileContents(fmt.Sprintf("%s/%s.json", constants.CLUSTER_RESOURCES_DIR, constants.CLUSTER_RESOURCES_NODES))
if err != nil {
return nil, errors.Wrap(err, "failed to get contents of nodes.json")
if _, ok := err.(*types.NotFoundError); !ok {
return nil, errors.Wrap(err, "failed to get contents of nodes.json")
}
if analyzer.IgnoreIfNoFiles {
return nil, nil
}
return &AnalyzeResult{
Title: a.Title(),
IconKey: "kubernetes_node_resources",
IconURI: "https://troubleshoot.sh/images/analyzer-icons/node-resources.svg?w=16&h=18",
IsWarn: true,
Message: "No node resources were collected, unable to analyze node resources",
}, nil
}
var nodes corev1.NodeList
pkg/analyze/node_resources_test.go
+55
View File
@@ -1,6 +1,7 @@
package analyzer
import (
"fmt"
"testing"
"github.com/stretchr/testify/assert"
@@ -10,6 +11,7 @@ import (
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
troubleshootv1beta2 "github.com/replicatedhq/troubleshoot/pkg/apis/troubleshoot/v1beta2"
"github.com/replicatedhq/troubleshoot/pkg/types"
)
func Test_compareNodeResourceConditionalToActual(t *testing.T) {
@@ -1684,3 +1686,56 @@ func Test_analyzeNodeResources(t *testing.T) {
})
}
}
func Test_analyzeNodeResources_NoFiles(t *testing.T) {
missingFile := func(name string) ([]byte, error) {
return nil, &types.NotFoundError{Name: name}
}
t.Run("emits warning when nodes.json is not collected", func(t *testing.T) {
req := require.New(t)
analyzer := &troubleshootv1beta2.NodeResources{
Outcomes: []*troubleshootv1beta2.Outcome{
{Pass: &troubleshootv1beta2.SingleOutcome{Message: "ok"}},
},
}
a := AnalyzeNodeResources{analyzer: analyzer}
got, err := a.Analyze(missingFile, nil)
req.NoError(err)
req.Len(got, 1)
req.True(got[0].IsWarn)
req.Equal("Node Resources", got[0].Title)
})
t.Run("ignoreIfNoFiles suppresses the warning", func(t *testing.T) {
req := require.New(t)
analyzer := &troubleshootv1beta2.NodeResources{
IgnoreIfNoFiles: true,
Outcomes: []*troubleshootv1beta2.Outcome{
{Pass: &troubleshootv1beta2.SingleOutcome{Message: "ok"}},
},
}
a := AnalyzeNodeResources{analyzer: analyzer}
got, err := a.Analyze(missingFile, nil)
req.NoError(err)
req.Empty(got)
})
t.Run("non-NotFound errors are propagated", func(t *testing.T) {
req := require.New(t)
analyzer := &troubleshootv1beta2.NodeResources{
IgnoreIfNoFiles: true,
Outcomes: []*troubleshootv1beta2.Outcome{
{Pass: &troubleshootv1beta2.SingleOutcome{Message: "ok"}},
},
}
ioErr := func(string) ([]byte, error) {
return nil, fmt.Errorf("permission denied")
}
a := AnalyzeNodeResources{analyzer: analyzer}
got, err := a.Analyze(ioErr, nil)
req.Error(err)
req.Nil(got)
req.Contains(err.Error(), "permission denied")
})
}
pkg/apis/troubleshoot/v1beta2/analyzer_shared.go
+4 -3
View File
@@ -130,9 +130,10 @@ type Distribution struct {
}
type NodeResources struct {
AnalyzeMeta `json:",inline" yaml:",inline"`
Outcomes []*Outcome `json:"outcomes" yaml:"outcomes"`
Filters *NodeResourceFilters `json:"filters,omitempty" yaml:"filters,omitempty"`
AnalyzeMeta `json:",inline" yaml:",inline"`
Outcomes []*Outcome `json:"outcomes" yaml:"outcomes"`
Filters *NodeResourceFilters `json:"filters,omitempty" yaml:"filters,omitempty"`
IgnoreIfNoFiles bool `json:"ignoreIfNoFiles,omitempty" yaml:"ignoreIfNoFiles,omitempty"`
}
type NodeResourceFilters struct {
schemas/analyzer-troubleshoot-v1beta2.json
+3
View File
@@ -1836,6 +1836,9 @@
"exclude": {
"oneOf": [{"type": "string"},{"type": "boolean"}]
},
"ignoreIfNoFiles": {
"type": "boolean"
},
"filters": {
"type": "object",
"properties": {
schemas/preflight-troubleshoot-v1beta2.json
+3
View File
@@ -1836,6 +1836,9 @@
"exclude": {
"oneOf": [{"type": "string"},{"type": "boolean"}]
},
"ignoreIfNoFiles": {
"type": "boolean"
},
"filters": {
"type": "object",
"properties": {
schemas/supportbundle-troubleshoot-v1beta2.json
+3
View File
@@ -1882,6 +1882,9 @@
"exclude": {
"oneOf": [{"type": "string"},{"type": "boolean"}]
},
"ignoreIfNoFiles": {
"type": "boolean"
},
"filters": {
"type": "object",
"properties": {
test/e2e/support-bundle/node_resources_no_files_e2e_test.go
+101
View File
@@ -0,0 +1,101 @@
package e2e
import (
"bytes"
"context"
"encoding/json"
"fmt"
"os"
"os/exec"
"testing"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
"sigs.k8s.io/e2e-framework/pkg/envconf"
"sigs.k8s.io/e2e-framework/pkg/features"
)
// TestNodeResourcesNoFiles verifies the behavior of the nodeResources
// analyzer when cluster-resources/nodes.json is not present in the
// bundle. The spec excludes the clusterResources collector so nodes.json
// is never written, then runs `support-bundle analyze` against the
// generated bundle.
//
// Expected outcomes:
// - "warn-default": warn outcome (default behavior)
// - "warn-explicit-false": warn outcome (ignoreIfNoFiles: false)
// - "ignored": no result (ignoreIfNoFiles: true)
func TestNodeResourcesNoFiles(t *testing.T) {
feature := features.New("Node Resources No Files").
Assess("warns per analyzer when nodes.json is missing and respects ignoreIfNoFiles", func(ctx context.Context, t *testing.T, c *envconf.Config) context.Context {
supportBundleName := "node-resources-no-files-test"
tarPath := fmt.Sprintf("%s.tar.gz", supportBundleName)
specPath := "spec/nodeResourcesNoFiles.yaml"
var collectOut bytes.Buffer
collectCmd := exec.CommandContext(ctx, sbBinary(), specPath,
"--interactive=false",
fmt.Sprintf("-o=%s", supportBundleName),
)
collectCmd.Stdout = &collectOut
collectCmd.Stderr = &collectOut
require.NoErrorf(t, collectCmd.Run(), "support-bundle collect failed: %s", collectOut.String())
defer func() {
if err := os.Remove(tarPath); err != nil {
t.Logf("Error removing %s: %v", tarPath, err)
}
}()
// Sanity check: nodes.json should NOT have been collected.
_, err := readFileFromTar(tarPath, fmt.Sprintf("%s/cluster-resources/nodes.json", supportBundleName))
require.Error(t, err, "nodes.json should not exist in the bundle")
var analyzeOut bytes.Buffer
var analyzeErr bytes.Buffer
analyzeCmd := exec.CommandContext(ctx, sbBinary(), "analyze",
"--bundle", tarPath,
"--output", "json",
specPath,
)
analyzeCmd.Stdout = &analyzeOut
analyzeCmd.Stderr = &analyzeErr
require.NoErrorf(t, analyzeCmd.Run(), "support-bundle analyze failed: %s", analyzeErr.String())
type analyzeResult struct {
IsPass bool `json:"IsPass"`
IsFail bool `json:"IsFail"`
IsWarn bool `json:"IsWarn"`
Title string `json:"Title"`
Message string `json:"Message"`
}
var results []analyzeResult
require.NoError(t, json.Unmarshal(analyzeOut.Bytes(), &results), "analyzer JSON output: %s", analyzeOut.String())
byTitle := map[string]analyzeResult{}
for _, r := range results {
byTitle[r.Title] = r
}
// Two warns, no result for the suppressed entry.
assert.Len(t, results, 2, "expected exactly two analyzer results, got %d: %s", len(results), analyzeOut.String())
for _, title := range []string{"warn-default", "warn-explicit-false"} {
r, ok := byTitle[title]
if !assert.Truef(t, ok, "expected an analyzer result with title %q", title) {
continue
}
assert.Truef(t, r.IsWarn, "%q: expected IsWarn=true", title)
assert.Falsef(t, r.IsFail, "%q: expected IsFail=false", title)
assert.Falsef(t, r.IsPass, "%q: expected IsPass=false", title)
assert.Containsf(t, r.Message, "No node resources were collected", "%q: unexpected message %q", title, r.Message)
}
_, ignored := byTitle["ignored"]
assert.Falsef(t, ignored, "analyzer with ignoreIfNoFiles: true should have produced no result")
return ctx
}).Feature()
testenv.Test(t, feature)
}
test/e2e/support-bundle/spec/nodeResourcesNoFiles.yaml
+37
View File
@@ -0,0 +1,37 @@
apiVersion: troubleshoot.sh/v1beta2
kind: SupportBundle
metadata:
name: node-resources-no-files-test
spec:
collectors:
- clusterResources:
exclude: true
- clusterInfo:
exclude: true
analyzers:
- nodeResources:
checkName: warn-default
outcomes:
- fail:
when: "count() < 3"
message: This application requires at least 3 nodes
- pass:
message: This cluster has enough nodes
- nodeResources:
checkName: warn-explicit-false
ignoreIfNoFiles: false
outcomes:
- fail:
when: "min(memoryCapacity) < 16Gi"
message: All nodes must have at least 16Gi of memory
- pass:
message: All nodes have at least 16Gi of memory
- nodeResources:
checkName: ignored
ignoreIfNoFiles: true
outcomes:
- fail:
when: "count() < 3"
message: This application requires at least 3 nodes
- pass:
message: This cluster has enough nodes