github/containerd
4
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 2 Wiki Activity

Merge pull request #4538 from fuweid/update-shim-cleanup

Browse Source 
runtime/v2: cleanup dead shim before delete bundle
This commit is contained in:
Phil Estes
2020-09-21 13:32:40 -04:00
committed by GitHub
parent 438c87b8e0 4b05d03903
commit 68d97331be
4 changed files with 44 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
runtime/v2/manager.go
View File

@@ -138,12 +138,8 @@ func (m *TaskManager) Create(ctx context.Context, id string, opts runtime.Create
b := shimBinary(ctx, bundle, opts.Runtime, m.containerdAddress, m.containerdTTRPCAddress, m.events, m.tasks)
shim, err := b.Start(ctx, topts, func() {
log.G(ctx).WithField("id", id).Info("shim disconnected")
_, err := m.tasks.Get(ctx, id)
if err != nil {
// Task was never started or was already successfully deleted
return
}
cleanupAfterDeadShim(context.Background(), id, ns, m.events, b)
cleanupAfterDeadShim(context.Background(), id, ns, m.tasks, m.events, b)
// Remove self from the runtime task list. Even though the cleanupAfterDeadShim()
// would publish taskExit event, but the shim.Delete() would always failed with ttrpc
// disconnect and there is no chance to remove this dead task from runtime task lists.
@@ -266,17 +262,13 @@ func (m *TaskManager) loadTasks(ctx context.Context) error {
binaryCall := shimBinary(ctx, bundle, container.Runtime.Name, m.containerdAddress, m.containerdTTRPCAddress, m.events, m.tasks)
shim, err := loadShim(ctx, bundle, m.events, m.tasks, func() {
log.G(ctx).WithField("id", id).Info("shim disconnected")
_, err := m.tasks.Get(ctx, id)
if err != nil {
// Task was never started or was already successfully deleted
return
}
cleanupAfterDeadShim(context.Background(), id, ns, m.events, binaryCall)
cleanupAfterDeadShim(context.Background(), id, ns, m.tasks, m.events, binaryCall)
// Remove self from the runtime task list.
m.tasks.Delete(ctx, id)
})
if err != nil {
cleanupAfterDeadShim(ctx, id, ns, m.events, binaryCall)
cleanupAfterDeadShim(ctx, id, ns, m.tasks, m.events, binaryCall)
continue
}
m.tasks.Add(ctx, shim)

16
runtime/v2/shim.go
View File

@@ -121,7 +121,7 @@ func loadShim(ctx context.Context, bundle *Bundle, events *exchange.Exchange, rt
return s, nil
}
func cleanupAfterDeadShim(ctx context.Context, id, ns string, events *exchange.Exchange, binaryCall *binary) {
func cleanupAfterDeadShim(ctx context.Context, id, ns string, rt *runtime.TaskList, events *exchange.Exchange, binaryCall *binary) {
ctx = namespaces.WithNamespace(ctx, ns)
ctx, cancel := timeout.WithContext(ctx, cleanupTimeout)
defer cancel()
@@ -138,6 +138,12 @@ func cleanupAfterDeadShim(ctx context.Context, id, ns string, events *exchange.E
}).Warn("failed to clean up after shim disconnected")
}
if _, err := rt.Get(ctx, id); err != nil {
// Task was never started or was already successfully deleted
// No need to publish events
return
}
var (
pid uint32
exitStatus uint32
@@ -234,13 +240,15 @@ func (s *shim) Delete(ctx context.Context) (*runtime.Exit, error) {
}
}
}
// remove self from the runtime task list
// this seems dirty but it cleans up the API across runtimes, tasks, and the service
s.rtTasks.Delete(ctx, s.ID())
if err := s.waitShutdown(ctx); err != nil {
log.G(ctx).WithField("id", s.ID()).WithError(err).Error("failed to shutdown shim")
}
s.Close()
s.client.UserOnCloseWait(ctx)
// remove self from the runtime task list
// this seems dirty but it cleans up the API across runtimes, tasks, and the service
s.rtTasks.Delete(ctx, s.ID())
if err := s.bundle.Delete(); err != nil {
log.G(ctx).WithField("id", s.ID()).WithError(err).Error("failed to delete bundle")
}