Skip to content
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

Jump to bottom

WRKLDS-1012: oc adm must-gather: pull gather container logs #1641

Merged
merged 2 commits into from
Jan 19, 2024
Merged

WRKLDS-1012: oc adm must-gather: pull gather container logs #1641

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
6348e4a
oc adm must-gather: pull gather container logs
ingvagabund Dec 19, 2023
7d23cbb
Write oc adm must-gather logs into a file
ingvagabund Dec 19, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
106 changes: 84 additions & 22 deletions pkg/cli/admin/mustgather/mustgather.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,6 +47,10 @@ import (
ocmdhelpers "github.com/openshift/oc/pkg/helpers/cmd"
)

const (
gatherContainerName = "gather"
)

var (
mustGatherLong = templates.LongDesc(`
Launch a pod to gather debugging information.
Expand Down Expand Up @@ -77,13 +81,13 @@ var (

volumeUsageCheckerScript = `
echo "volume percentage checker started....."
while true; do
while true; do
disk_usage=$(du -s "%s" | awk '{print $1}')
disk_space=$(df -P "%s" | awk 'NR==2 {print $2}')
usage_percentage=$(( (disk_usage * 100) / disk_space ))
echo "volume usage percentage $usage_percentage"
if [ "$usage_percentage" -gt "%d" ]; then
echo "Disk usage exceeds the volume percentage of %d for mounted directory. Exiting..."
echo "volume usage percentage $usage_percentage"
if [ "$usage_percentage" -gt "%d" ]; then
echo "Disk usage exceeds the volume percentage of %d for mounted directory. Exiting..."
# kill gathering process in gather container to prevent disk to use more.
pkill --signal SIGKILL -f %s
exit 1
Expand Down Expand Up @@ -125,14 +129,15 @@ func NewMustGatherCommand(f kcmdutil.Factory, streams genericiooptions.IOStreams
}

func NewMustGatherOptions(streams genericiooptions.IOStreams) *MustGatherOptions {
return &MustGatherOptions{
opts := &MustGatherOptions{
SourceDir: "/must-gather/",
IOStreams: streams,
LogOut: newPrefixWriter(streams.Out, "[must-gather ] OUT"),
RawOut: streams.Out,
Timeout: 10 * time.Minute,
VolumePercentage: 30,
}
opts.LogOut = opts.newPrefixWriter(streams.Out, "[must-gather ] OUT", false)
opts.RawOut = opts.newPrefixWriter(streams.Out, "", false)
return opts
}

func (o *MustGatherOptions) Complete(f kcmdutil.Factory, cmd *cobra.Command, args []string) error {
Expand Down Expand Up @@ -206,6 +211,7 @@ func (o *MustGatherOptions) completeImages() error {
o.Images = append(o.Images, image)
}
o.log("Using must-gather plug-in image: %s", strings.Join(o.Images, ", "))

return nil
}

Expand Down Expand Up @@ -275,6 +281,9 @@ type MustGatherOptions struct {
LogOut io.Writer
// RawOut is used for printing information we're looking to have copy/pasted into bugs
RawOut io.Writer

LogWriter *os.File
LogWriterMux sync.Mutex
}

func (o *MustGatherOptions) Validate() error {
Expand Down Expand Up @@ -316,6 +325,27 @@ func (o *MustGatherOptions) Validate() error {
func (o *MustGatherOptions) Run() error {
var errs []error

if err := os.MkdirAll(o.DestDir, os.ModePerm); err != nil {
// ensure the errors bubble up to BackupGathering method for display
errs = []error{err}
return err
}

f, err := os.Create(path.Join(o.DestDir, "must-gather.logs"))
if err != nil {
fmt.Fprintf(os.Stderr, "Unable to write must-gather logs: %v. It is possible the destination directory has not been created yet due to early termination\n", err)
} else {
o.LogWriter = f
o.LogWriterMux.Lock()
// gets invoked in Complete step before must-gather.logs is created
o.LogWriter.WriteString(fmt.Sprintf("[must-gather ] OUT Using must-gather plug-in image: %s\n", strings.Join(o.Images, ", ")))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What if user passes --all-images flag?. Are we handling it also?.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We are. https://github.com/openshift/oc/blob/master/pkg/cli/admin/mustgather/mustgather.go#L208 populates o.Images which includes all images. Changes in #1633 extend completeImages() which is invoked before this line.

o.LogWriterMux.Unlock()

defer func() {
o.LogWriter.Close()
}()
}

// print at both the beginning and at the end. This information is important enough to be in both spots.
o.PrintBasicClusterState(context.TODO())
defer func() {
Expand Down Expand Up @@ -413,11 +443,6 @@ func (o *MustGatherOptions) Run() error {
}

// log timestamps...
if err := os.MkdirAll(o.DestDir, os.ModePerm); err != nil {
// ensure the errors bubble up to BackupGathering method for display
errs = []error{err}
return err
}
if err := o.logTimestamp(); err != nil {
// ensure the errors bubble up to BackupGathering method for display
errs = []error{err}
Expand Down Expand Up @@ -445,7 +470,7 @@ func (o *MustGatherOptions) Run() error {
}()
}

log := newPodOutLogger(o.Out, pod.Name)
log := o.newPodOutLogger(o.Out, pod.Name)

// wait for gather container to be running (gather is running)
if err := o.waitForGatherContainerRunning(pod); err != nil {
Expand Down Expand Up @@ -509,8 +534,8 @@ func (o *MustGatherOptions) Run() error {
return errors.NewAggregate(errs)
}

func newPodOutLogger(out io.Writer, podName string) func(string, ...interface{}) {
writer := newPrefixWriter(out, fmt.Sprintf("[%s] OUT", podName))
func (o *MustGatherOptions) newPodOutLogger(out io.Writer, podName string) func(string, ...interface{}) {
writer := o.newPrefixWriter(out, fmt.Sprintf("[%s] OUT", podName), false)
return func(format string, a ...interface{}) {
fmt.Fprintf(writer, format+"\n", a...)
}
Expand All @@ -531,7 +556,7 @@ func (o *MustGatherOptions) logTimestamp() error {

func (o *MustGatherOptions) copyFilesFromPod(pod *corev1.Pod) error {
streams := o.IOStreams
streams.Out = newPrefixWriter(streams.Out, fmt.Sprintf("[%s] OUT", pod.Name))
streams.Out = o.newPrefixWriter(streams.Out, fmt.Sprintf("[%s] OUT", pod.Name), false)
imageFolder := regexp.MustCompile("[^A-Za-z0-9]+").ReplaceAllString(pod.Status.ContainerStatuses[0].ImageID, "-")
var destDir string
if o.NodeSelector != "" {
Expand All @@ -542,6 +567,33 @@ func (o *MustGatherOptions) copyFilesFromPod(pod *corev1.Pod) error {
if err := os.MkdirAll(destDir, os.ModePerm); err != nil {
return err
}

var errs []error

// get must-gather gather container logs
if err := func() error {
dest, err := os.OpenFile(path.Join(destDir, "/gather.logs"), os.O_RDWR|os.O_CREATE|os.O_TRUNC, 0755)
if err != nil {
return err
}
defer dest.Close()

logOptions := &corev1.PodLogOptions{
Container: gatherContainerName,
Timestamps: true,
}
readCloser, err := o.Client.CoreV1().Pods(pod.Namespace).GetLogs(pod.Name, logOptions).Stream(context.TODO())
if err != nil {
return err
}
defer readCloser.Close()

_, err = io.Copy(dest, readCloser)
return err
}(); err != nil {
errs = append(errs, err)
}

rsyncOptions := &rsync.RsyncOptions{
Namespace: pod.Namespace,
Source: &rsync.PathSpec{PodName: pod.Name, Path: path.Clean(o.SourceDir) + "/"},
Expand All @@ -559,8 +611,9 @@ func (o *MustGatherOptions) copyFilesFromPod(pod *corev1.Pod) error {
klog.V(4).Infof("re-trying rsync after initial failure %v", err)
// re-try copying data before letting it go
err = rsyncOptions.RunRsync()
errs = append(errs, err)
}
return err
return errors.NewAggregate(errs)
}

func (o *MustGatherOptions) getGatherContainerLogs(pod *corev1.Pod) error {
Expand All @@ -577,7 +630,7 @@ func (o *MustGatherOptions) getGatherContainerLogs(pod *corev1.Pod) error {
Object: pod,
ConsumeRequestFn: logs.DefaultConsumeRequest,
LogsForObject: polymorphichelpers.LogsForObjectFn,
IOStreams: genericiooptions.IOStreams{Out: newPrefixWriter(o.Out, fmt.Sprintf("[%s] POD", pod.Name))},
IOStreams: genericiooptions.IOStreams{Out: o.newPrefixWriter(o.Out, fmt.Sprintf("[%s] POD", pod.Name), true)},
}

for {
Expand All @@ -597,12 +650,21 @@ func (o *MustGatherOptions) getGatherContainerLogs(pod *corev1.Pod) error {
}
}

func newPrefixWriter(out io.Writer, prefix string) io.Writer {
func (o *MustGatherOptions) newPrefixWriter(out io.Writer, prefix string, ignoreFileWriter bool) io.Writer {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In which circumstance we should pass ignoreFileWriter to true?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

E.g. when logs are printed for individual must-gather running container logs.

reader, writer := io.Pipe()
scanner := bufio.NewScanner(reader)
if prefix != "" {
prefix = prefix + " "
}
go func() {
for scanner.Scan() {
fmt.Fprintf(out, "%s %s\n", prefix, scanner.Text())
text := scanner.Text()
if !ignoreFileWriter && o.LogWriter != nil {
o.LogWriterMux.Lock()
o.LogWriter.WriteString(fmt.Sprintf("%s%s\n", prefix, text))
o.LogWriterMux.Unlock()
}
fmt.Fprintf(out, "%s%s\n", prefix, text)
}
}()
return writer
Expand All @@ -625,7 +687,7 @@ func (o *MustGatherOptions) isGatherDone(pod *corev1.Pod) (bool, error) {
}
var state *corev1.ContainerState
for _, cstate := range pod.Status.ContainerStatuses {
if cstate.Name == "gather" {
if cstate.Name == gatherContainerName {
state = &cstate.State
break
}
Expand Down Expand Up @@ -803,7 +865,7 @@ func (o *MustGatherOptions) newPod(node, image string, hasMaster bool) *corev1.P
},
Containers: []corev1.Container{
{
Name: "gather",
Name: gatherContainerName,
Image: image,
ImagePullPolicy: corev1.PullIfNotPresent,
// always force disk flush to ensure that all data gathered is accessible in the copy container
Expand Down
8 changes: 4 additions & 4 deletions pkg/cli/admin/mustgather/mustgather_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,18 +114,18 @@ func TestGetNamespace(t *testing.T) {
t.Parallel()

for name, tc := range map[string]struct {
Options MustGatherOptions
Options *MustGatherOptions
ShouldBeRetained bool
ShouldFail bool
}{
"no namespace given": {
Options: MustGatherOptions{
Options: &MustGatherOptions{
Client: fake.NewSimpleClientset(),
},
ShouldBeRetained: false,
},
"namespace given": {
Options: MustGatherOptions{
Options: &MustGatherOptions{
Client: fake.NewSimpleClientset(
&corev1.Namespace{
ObjectMeta: metav1.ObjectMeta{
Expand All @@ -138,7 +138,7 @@ func TestGetNamespace(t *testing.T) {
ShouldBeRetained: true,
},
"namespace given, but does not exist": {
Options: MustGatherOptions{
Options: &MustGatherOptions{
Client: fake.NewSimpleClientset(),
RunNamespace: "test-namespace",
},
Expand Down