Thanks to visit codestin.com
Credit goes to github.com

Skip to content

[release-1.23] Fix vm containers could not restore after CRI-O restart #5633

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

Merged
openshift-merge-robot merged 2 commits into from
Feb 14, 2022
Merged

[release-1.23] Fix vm containers could not restore after CRI-O restart #5633

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
c9b4eb8
Fix vm containers couldn't restore after CRI-O restart
sleepymole Jan 29, 2022
e19f812
Reuse createContainerIO in CreateContainer
sleepymole Jan 29, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 9 additions & 2 deletions internal/oci/oci.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -241,10 +241,17 @@ func (r *Runtime) RuntimeImpl(c *Container) (RuntimeImpl, error) {
r.runtimeImplMapMutex.RLock()
impl, ok := r.runtimeImplMap[c.ID()]
r.runtimeImplMapMutex.RUnlock()
if !ok {
return r.newRuntimeImpl(c)
if ok {
return impl, nil
}

impl, err := r.newRuntimeImpl(c)
if err != nil {
return nil, err
}
r.runtimeImplMapMutex.Lock()
r.runtimeImplMap[c.ID()] = impl
r.runtimeImplMapMutex.Unlock()
return impl, nil
}

Expand Down
158 changes: 115 additions & 43 deletions internal/oci/runtime_vm.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@ package oci
import (
"bytes"
"io"
"io/ioutil"
"os"
"path/filepath"
"strconv"
Expand All @@ -13,6 +14,7 @@ import (

cgroups "github.com/containerd/cgroups/stats/v1"
tasktypes "github.com/containerd/containerd/api/types/task"
ctrio "github.com/containerd/containerd/cio"
"github.com/containerd/containerd/namespaces"
client "github.com/containerd/containerd/runtime/v2/shim"
"github.com/containerd/containerd/runtime/v2/task"
Expand Down Expand Up @@ -121,52 +123,14 @@ func (r *runtimeVM) CreateContainer(ctx context.Context, c *Container, cgroupPar
return err
}

// Create IO fifos
containerIO, err := cio.NewContainerIO(c.ID(),
cio.WithNewFIFOs(r.fifoDir, c.terminal, c.stdin))
containerIO, err := r.createContainerIO(ctx, c, cio.WithNewFIFOs(r.fifoDir, c.terminal, c.stdin))
if err != nil {
return err
}

defer func() {
if retErr != nil {
containerIO.Close()
}
}()

f, err := os.OpenFile(c.LogPath(), os.O_WRONLY|os.O_APPEND|os.O_CREATE, 0o600)
if err != nil {
return err
}

var stdoutCh, stderrCh <-chan struct{}
wc := cioutil.NewSerialWriteCloser(f)
stdout, stdoutCh := cio.NewCRILogger(c.LogPath(), wc, cio.Stdout, -1)
stderr, stderrCh := cio.NewCRILogger(c.LogPath(), wc, cio.Stderr, -1)

go func() {
if stdoutCh != nil {
<-stdoutCh
}
if stderrCh != nil {
<-stderrCh
}
log.Debugf(ctx, "Finish redirecting log file %q, closing it", c.LogPath())
f.Close()
}()

containerIO.AddOutput(c.LogPath(), stdout, stderr)
containerIO.Pipe()

r.Lock()
r.ctrs[c.ID()] = containerInfo{
cio: containerIO,
}
r.Unlock()

defer func() {
if retErr != nil {
log.Warnf(ctx, "Cleaning up container %s: %v", c.ID(), err)
log.Warnf(ctx, "Cleaning up container %s: %v", c.ID(), retErr)
if cleanupErr := r.deleteContainer(c, true); cleanupErr != nil {
log.Infof(ctx, "DeleteContainer failed for container %s: %v", c.ID(), cleanupErr)
}
Expand Down Expand Up @@ -664,10 +628,24 @@ func (r *runtimeVM) updateContainerStatus(ctx context.Context, c *Container) err
log.Debugf(ctx, "RuntimeVM.updateContainerStatus() start")
defer log.Debugf(ctx, "RuntimeVM.updateContainerStatus() end")

// This can happen on restore, for example if we switch the runtime type
// for a container from "oci" to "vm" for the same runtime.
// This can happen on restore. We need to read shim address from the bundle path.
// And then connect to the existing gRPC server with this address.
if r.task == nil {
return errors.New("runtime not correctly setup")
addressPath := filepath.Join(c.BundlePath(), "address")
data, err := ioutil.ReadFile(addressPath)
if err != nil {
log.Warnf(ctx, "Failed to read shim address: %v", err)
return errors.New("runtime not correctly setup")
}
address := strings.TrimSpace(string(data))
conn, err := client.Connect(address, client.AnonDialer)
if err != nil {
return err
}
options := ttrpc.WithOnClose(func() { conn.Close() })
cl := ttrpc.NewClient(conn, options)
r.client = cl
r.task = task.NewTaskClient(cl)
}

response, err := r.task.State(r.ctx, &task.StateRequest{
Expand All @@ -680,6 +658,10 @@ func (r *runtimeVM) updateContainerStatus(ctx context.Context, c *Container) err
return errdefs.ErrNotFound
}

if err = r.restoreContainerIO(ctx, c, response); err != nil {
return errors.Wrapf(err, "failed to restore container io")
}

status := c.state.Status
switch response.Status {
case tasktypes.StatusCreated:
Expand Down Expand Up @@ -713,6 +695,96 @@ func (r *runtimeVM) updateContainerStatus(ctx context.Context, c *Container) err
return nil
}

func (r *runtimeVM) restoreContainerIO(ctx context.Context, c *Container, state *task.StateResponse) error {
r.Lock()
_, ok := r.ctrs[c.ID()]
if ok {
r.Unlock()
return nil
}
r.Unlock()

cioCfg := ctrio.Config{
Terminal: state.Terminal,
Stdin: state.Stdin,
Stdout: state.Stdout,
Stderr: state.Stderr,
}
// The existing fifos is created by NewFIFOSetInDir. stdin, stdout, stderr should exist
// in a same temporary directory under r.fifoDir. crio is responsible for removing these
// files after container io is closed.
var iofiles []string
if cioCfg.Stdin != "" {
iofiles = append(iofiles, cioCfg.Stdin)
}
if cioCfg.Stdout != "" {
iofiles = append(iofiles, cioCfg.Stdout)
}
if cioCfg.Stderr != "" {
iofiles = append(iofiles, cioCfg.Stderr)
}
closer := func() error {
for _, f := range iofiles {
if err := os.Remove(f); err != nil {
return err
}
}
// Also try to remove the parent dir if it is empty.
for _, f := range iofiles {
_ = os.Remove(filepath.Dir(f))
}
return nil
}
_, err := r.createContainerIO(ctx, c, cio.WithFIFOs(ctrio.NewFIFOSet(cioCfg, closer)))
return err
}

func (r *runtimeVM) createContainerIO(ctx context.Context, c *Container, cioOpts ...cio.ContainerIOOpts) (_ *cio.ContainerIO, retErr error) {
// Create IO fifos
containerIO, err := cio.NewContainerIO(c.ID(), cioOpts...)
if err != nil {
return nil, err
}

defer func() {
if retErr != nil {
containerIO.Close()
}
}()

f, err := os.OpenFile(c.LogPath(), os.O_WRONLY|os.O_APPEND|os.O_CREATE, 0o600)
if err != nil {
return nil, err
}

var stdoutCh, stderrCh <-chan struct{}
wc := cioutil.NewSerialWriteCloser(f)
stdout, stdoutCh := cio.NewCRILogger(c.LogPath(), wc, cio.Stdout, -1)
stderr, stderrCh := cio.NewCRILogger(c.LogPath(), wc, cio.Stderr, -1)

go func() {
if stdoutCh != nil {
<-stdoutCh
}
if stderrCh != nil {
<-stderrCh
}
log.Debugf(ctx, "Finish redirecting log file %q, closing it", c.LogPath())
f.Close()
}()

containerIO.AddOutput(c.LogPath(), stdout, stderr)
containerIO.Pipe()

r.Lock()
r.ctrs[c.ID()] = containerInfo{
cio: containerIO,
}
r.Unlock()

return containerIO, nil
}

// PauseContainer pauses a container.
func (r *runtimeVM) PauseContainer(ctx context.Context, c *Container) error {
log.Debugf(ctx, "RuntimeVM.PauseContainer() start")
Expand Down