Fix exit event handling in shim

Could issues where when exec processes fail the wait block is not
released.

Second, you could not dump stacks if the reaper loop locks up.

Third, the publisher was not waiting on the correct pid.

Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
This commit is contained in:
Michael Crosby 2017-11-28 14:17:22 -05:00
parent 723f37d846
commit 74b3cb3391
4 changed files with 18 additions and 7 deletions

View File

@ -11,6 +11,7 @@ import (
"net" "net"
"os" "os"
"os/exec" "os/exec"
"os/signal"
"runtime" "runtime"
"strings" "strings"
"sync" "sync"
@ -80,6 +81,9 @@ func executeShim() error {
if err != nil { if err != nil {
return err return err
} }
dump := make(chan os.Signal, 32)
signal.Notify(dump, syscall.SIGUSR1)
path, err := os.Getwd() path, err := os.Getwd()
if err != nil { if err != nil {
return err return err
@ -111,6 +115,11 @@ func executeShim() error {
"path": path, "path": path,
"namespace": namespaceFlag, "namespace": namespaceFlag,
}) })
go func() {
for range dump {
dumpStacks(logger)
}
}()
return handleSignals(logger, signals, server, sv) return handleSignals(logger, signals, server, sv)
} }
@ -171,8 +180,6 @@ func handleSignals(logger *logrus.Entry, signals chan os.Signal, server *ttrpc.S
sv.Delete(context.Background(), &ptypes.Empty{}) sv.Delete(context.Background(), &ptypes.Empty{})
close(done) close(done)
}) })
case unix.SIGUSR1:
dumpStacks(logger)
} }
} }
} }
@ -213,8 +220,11 @@ func (l *remoteEventsPublisher) Publish(ctx context.Context, topic string, event
if err != nil { if err != nil {
return err return err
} }
exit := <-c status, err := reaper.Default.Wait(cmd, c)
if exit.Status != 0 { if err != nil {
return err
}
if status != 0 {
return errors.New("failed to publish event") return errors.New("failed to publish event")
} }
return nil return nil

View File

@ -3,6 +3,7 @@ package main
import ( import (
"os" "os"
"os/signal" "os/signal"
"syscall"
"github.com/containerd/containerd/reaper" "github.com/containerd/containerd/reaper"
"github.com/containerd/containerd/sys" "github.com/containerd/containerd/sys"
@ -14,7 +15,7 @@ import (
// sub-reaper so that the container processes are reparented // sub-reaper so that the container processes are reparented
func setupSignals() (chan os.Signal, error) { func setupSignals() (chan os.Signal, error) {
signals := make(chan os.Signal, 2048) signals := make(chan os.Signal, 2048)
signal.Notify(signals) signal.Notify(signals, syscall.SIGTERM, syscall.SIGINT, syscall.SIGCHLD)
// make sure runc is setup to use the monitor // make sure runc is setup to use the monitor
// for waiting on processes // for waiting on processes
runc.Monitor = reaper.Default runc.Monitor = reaper.Default

View File

@ -143,6 +143,7 @@ func (e *execProcess) start(ctx context.Context) (err error) {
opts.ConsoleSocket = socket opts.ConsoleSocket = socket
} }
if err := e.parent.runtime.Exec(ctx, e.parent.id, e.spec, opts); err != nil { if err := e.parent.runtime.Exec(ctx, e.parent.id, e.spec, opts); err != nil {
close(e.waitBlock)
return e.parent.runtimeError(err, "OCI runtime exec failed") return e.parent.runtimeError(err, "OCI runtime exec failed")
} }
if e.stdio.Stdin != "" { if e.stdio.Stdin != "" {

View File

@ -148,7 +148,6 @@ func (s *Service) Delete(ctx context.Context, r *ptypes.Empty) (*shimapi.DeleteR
if p == nil { if p == nil {
return nil, errdefs.ToGRPCf(errdefs.ErrFailedPrecondition, "container must be created") return nil, errdefs.ToGRPCf(errdefs.ErrFailedPrecondition, "container must be created")
} }
if err := p.Delete(ctx); err != nil { if err := p.Delete(ctx); err != nil {
return nil, err return nil, err
} }
@ -480,7 +479,7 @@ func (s *Service) getContainerPids(ctx context.Context, id string) ([]uint32, er
func (s *Service) forward(publisher events.Publisher) { func (s *Service) forward(publisher events.Publisher) {
for e := range s.events { for e := range s.events {
if err := publisher.Publish(s.context, getTopic(s.context, e), e); err != nil { if err := publisher.Publish(s.context, getTopic(s.context, e), e); err != nil {
logrus.WithError(err).Error("post event") log.G(s.context).WithError(err).Error("post event")
} }
} }
} }