1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280
|
package kubernetes
import (
"bufio"
"context"
"errors"
"fmt"
"io"
"strconv"
"strings"
"time"
"github.com/sirupsen/logrus"
"golang.org/x/sync/errgroup"
"k8s.io/client-go/kubernetes"
restclient "k8s.io/client-go/rest"
)
type logStreamer interface {
Stream(offset int64, output io.Writer) error
fmt.Stringer
}
type kubernetesLogStreamer struct {
kubernetesLogProcessorPodConfig
client *kubernetes.Clientset
clientConfig *restclient.Config
executor RemoteExecutor
}
func (s *kubernetesLogStreamer) Stream(offset int64, output io.Writer) error {
exec := ExecOptions{
Namespace: s.namespace,
PodName: s.pod,
ContainerName: s.container,
Stdin: false,
Command: []string{
"gitlab-runner-helper",
"read-logs",
"--path",
s.logPath,
"--offset",
strconv.FormatInt(offset, 10),
"--wait-file-timeout",
s.waitLogFileTimeout.String(),
},
Out: output,
Err: output,
Executor: s.executor,
Client: s.client,
Config: s.clientConfig,
}
return exec.executeRequest()
}
func (s *kubernetesLogStreamer) String() string {
return fmt.Sprintf("%s/%s/%s:%s", s.namespace, s.pod, s.container, s.logPath)
}
type logProcessor interface {
// Process listens for log lines
// consumers must read from the channel until it's closed
// consumers are also notified in case of error through the error channel
Process(ctx context.Context) (<-chan string, <-chan error)
}
type backoffCalculator interface {
ForAttempt(attempt float64) time.Duration
}
// kubernetesLogProcessor processes the logs from a container and tries to reattach
// to the stream constantly, stopping only when the passed context is cancelled.
type kubernetesLogProcessor struct {
backoff backoffCalculator
logger logrus.FieldLogger
logStreamer logStreamer
logsOffset int64
}
type kubernetesLogProcessorPodConfig struct {
namespace string
pod string
container string
logPath string
waitLogFileTimeout time.Duration
}
func newKubernetesLogProcessor(
client *kubernetes.Clientset,
clientConfig *restclient.Config,
backoff backoffCalculator,
logger logrus.FieldLogger,
podCfg kubernetesLogProcessorPodConfig,
) *kubernetesLogProcessor {
logStreamer := &kubernetesLogStreamer{
kubernetesLogProcessorPodConfig: podCfg,
client: client,
clientConfig: clientConfig,
executor: new(DefaultRemoteExecutor),
}
return &kubernetesLogProcessor{
backoff: backoff,
logger: logger,
logStreamer: logStreamer,
}
}
func (l *kubernetesLogProcessor) Process(ctx context.Context) (<-chan string, <-chan error) {
outCh := make(chan string)
errCh := make(chan error)
go func() {
defer close(outCh)
defer close(errCh)
l.attach(ctx, outCh, errCh)
}()
return outCh, errCh
}
func (l *kubernetesLogProcessor) attach(ctx context.Context, outCh chan string, errCh chan error) {
var (
attempt float64 = -1
backoffDuration time.Duration
)
for {
// We do not exit because we need the processLogs goroutine still running.
// Once the error message is sent, a new step cleanup variables is started.
// As the pod is still running, the processLogs goroutine is not launched anymore.
// This is why, even though the error is sent to fail the ongoing step,
// we keep trying to reconnect to the output log, as a new one is created for variables cleanup.
attempt++
if attempt > 0 {
backoffDuration = l.backoff.ForAttempt(attempt)
l.logger.Debugln(fmt.Sprintf(
"Backing off reattaching log for %s for %s (attempt %f)",
l.logStreamer,
backoffDuration,
attempt,
))
}
select {
case <-ctx.Done():
l.logger.Debugln(fmt.Sprintf("Detaching from log... %v", ctx.Err()))
return
case <-time.After(backoffDuration):
err := l.processStream(ctx, outCh)
exitCode := getExitCode(err)
switch {
case exitCode == outputLogFileNotExistsExitCode:
// The cleanup variables step recreates a new output.log file
// where the shells.TrapCommandExitStatus is written.
// To not miss this line, we need to have the offset reset when we reconnect to the newly created log
l.logsOffset = 0
errCh <- fmt.Errorf("output log file deleted, cannot continue %w", err)
case err != nil:
l.logger.Warningln(fmt.Sprintf("Error %v. Retrying...", err))
default:
l.logger.Debug("processStream exited with no error")
}
}
}
}
func (l *kubernetesLogProcessor) processStream(ctx context.Context, outCh chan string) error {
reader, writer := io.Pipe()
defer func() {
_ = reader.Close()
_ = writer.Close()
}()
// Using errgroup.WithContext doesn't work here since if either one of the goroutines
// exits with a nil error, we can't signal the other one to exit
ctx, cancel := context.WithCancel(ctx)
var gr errgroup.Group
logsOffset := l.logsOffset
gr.Go(func() error {
defer cancel()
err := l.logStreamer.Stream(logsOffset, writer)
// prevent printing an error that the container exited
// when the context is already cancelled
if errors.Is(ctx.Err(), context.Canceled) {
return nil
}
if err != nil {
err = fmt.Errorf("streaming logs %s: %w", l.logStreamer, err)
}
return err
})
gr.Go(func() error {
defer cancel()
err := l.readLogs(ctx, reader, outCh)
if err != nil {
err = fmt.Errorf("reading logs %s: %w", l.logStreamer, err)
}
return err
})
return gr.Wait()
}
func (l *kubernetesLogProcessor) readLogs(ctx context.Context, logs io.Reader, outCh chan string) error {
logsScanner, linesCh := l.scan(ctx, logs)
for {
select {
case <-ctx.Done():
return nil
case line, more := <-linesCh:
if !more {
l.logger.Debug("No more data in linesCh")
return logsScanner.Err()
}
newLogsOffset, logLine := l.parseLogLine(line)
if newLogsOffset != -1 {
l.logsOffset = newLogsOffset
}
outCh <- logLine
}
}
}
func (l *kubernetesLogProcessor) scan(ctx context.Context, logs io.Reader) (*bufio.Scanner, <-chan string) {
logsScanner := bufio.NewScanner(logs)
linesCh := make(chan string)
go func() {
defer close(linesCh)
// This goroutine will exit when the calling method closes the logs stream or the context is cancelled
for logsScanner.Scan() {
select {
case <-ctx.Done():
return
case linesCh <- logsScanner.Text():
}
}
}()
return logsScanner, linesCh
}
// Each line starts with its bytes offset. We need this to resume the log from that point
// if we detach for some reason. The format is "10 log line continues as normal".
// The line doesn't include the new line character.
// Lines without offset are acceptable and return -1 for offset.
func (l *kubernetesLogProcessor) parseLogLine(line string) (int64, string) {
if line == "" {
return -1, ""
}
offsetIndex := strings.Index(line, " ")
if offsetIndex == -1 {
return -1, line
}
offset := line[:offsetIndex]
parsedOffset, err := strconv.ParseInt(offset, 10, 64)
if err != nil {
return -1, line
}
logLine := line[offsetIndex+1:]
return parsedOffset, logLine
}
|