Files
kata-containers/src/runtime/virtcontainers/api.go
Samuel Ortiz dc7e9bce73 virtcontainers: sandbox: Host cgroups partitioning
This is a simplification of the host cgroup handling by partitioning the
host cgroups into 2: A sandbox cgroup and an overhead cgroup.

The sandbox cgroup is always created and initialized. The overhead
cgroup is only available when sandbox_cgroup_only is unset, and is
unconstrained on all controllers. The goal of having an overhead cgroup
is to be more flexible on how we manage a pod overhead. Having such
cgroup will allow for setting a fixed overhead per pod, for a subset of
controllers, while at the same time not having the pod being accounted
for those resources.

When sandbox_cgroup_only is not set, we move all non vCPU threads
to the overhead cgroup and let them run unconstrained. When it is set,
all pod related processes and threads will run in the sandbox cgroup.

Signed-off-by: Samuel Ortiz <samuel.e.ortiz@protonmail.com>
2021-09-14 07:09:29 +02:00

167 lines
3.9 KiB
Go

// Copyright (c) 2016 Intel Corporation
//
// SPDX-License-Identifier: Apache-2.0
//
package virtcontainers
import (
"context"
"runtime"
"github.com/kata-containers/kata-containers/src/runtime/pkg/katautils/katatrace"
deviceApi "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/device/api"
deviceConfig "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/device/config"
"github.com/kata-containers/kata-containers/src/runtime/virtcontainers/pkg/cgroups"
"github.com/kata-containers/kata-containers/src/runtime/virtcontainers/pkg/compatoci"
vcTypes "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/pkg/types"
"github.com/sirupsen/logrus"
)
// apiTracingTags defines tags for the trace span
var apiTracingTags = map[string]string{
"source": "runtime",
"packages": "virtcontainers",
"subsystem": "api",
}
func init() {
runtime.LockOSThread()
}
var virtLog = logrus.WithField("source", "virtcontainers")
// SetLogger sets the logger for virtcontainers package.
func SetLogger(ctx context.Context, logger *logrus.Entry) {
fields := virtLog.Data
virtLog = logger.WithFields(fields)
deviceApi.SetLogger(virtLog)
compatoci.SetLogger(virtLog)
deviceConfig.SetLogger(virtLog)
cgroups.SetLogger(virtLog)
}
// CreateSandbox is the virtcontainers sandbox creation entry point.
// CreateSandbox creates a sandbox and its containers. It does not start them.
func CreateSandbox(ctx context.Context, sandboxConfig SandboxConfig, factory Factory) (VCSandbox, error) {
span, ctx := katatrace.Trace(ctx, virtLog, "CreateSandbox", apiTracingTags)
defer span.End()
s, err := createSandboxFromConfig(ctx, sandboxConfig, factory)
return s, err
}
func createSandboxFromConfig(ctx context.Context, sandboxConfig SandboxConfig, factory Factory) (_ *Sandbox, err error) {
span, ctx := katatrace.Trace(ctx, virtLog, "createSandboxFromConfig", apiTracingTags)
defer span.End()
// Create the sandbox.
s, err := createSandbox(ctx, sandboxConfig, factory)
if err != nil {
return nil, err
}
// cleanup sandbox resources in case of any failure
defer func() {
if err != nil {
s.Delete(ctx)
}
}()
// Create the sandbox network
if err = s.createNetwork(ctx); err != nil {
return nil, err
}
// network rollback
defer func() {
if err != nil {
s.removeNetwork(ctx)
}
}()
// Set the sandbox host cgroups.
if err := s.setupCgroups(); err != nil {
return nil, err
}
// Start the VM
if err = s.startVM(ctx); err != nil {
return nil, err
}
// rollback to stop VM if error occurs
defer func() {
if err != nil {
s.stopVM(ctx)
}
}()
s.postCreatedNetwork(ctx)
if err = s.getAndStoreGuestDetails(ctx); err != nil {
return nil, err
}
// Create Containers
if err = s.createContainers(ctx); err != nil {
return nil, err
}
return s, nil
}
// CleanupContainer is used by shimv2 to stop and delete a container exclusively, once there is no container
// in the sandbox left, do stop the sandbox and delete it. Those serial operations will be done exclusively by
// locking the sandbox.
func CleanupContainer(ctx context.Context, sandboxID, containerID string, force bool) error {
span, ctx := katatrace.Trace(ctx, virtLog, "CleanupContainer", apiTracingTags)
defer span.End()
if sandboxID == "" {
return vcTypes.ErrNeedSandboxID
}
if containerID == "" {
return vcTypes.ErrNeedContainerID
}
unlock, err := rwLockSandbox(sandboxID)
if err != nil {
return err
}
defer unlock()
s, err := fetchSandbox(ctx, sandboxID)
if err != nil {
return err
}
defer s.Release(ctx)
_, err = s.StopContainer(ctx, containerID, force)
if err != nil && !force {
return err
}
_, err = s.DeleteContainer(ctx, containerID)
if err != nil && !force {
return err
}
if len(s.GetAllContainers()) > 0 {
return nil
}
if err = s.Stop(ctx, force); err != nil && !force {
return err
}
if err = s.Delete(ctx); err != nil {
return err
}
return nil
}