Skip to content

Identify permanent errors upon reading checkpoint #375

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

Closed
wants to merge 1 commit into from
Closed

Identify permanent errors upon reading checkpoint #375

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
17 changes: 17 additions & 0 deletions cmd/compute-domain-kubelet-plugin/device_state.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,8 @@ package main

import (
"context"
"encoding/json"
"errors"
"fmt"
"slices"
"sync"
Expand All @@ -27,6 +29,7 @@ import (
"k8s.io/dynamic-resource-allocation/kubeletplugin"
"k8s.io/klog/v2"
"k8s.io/kubernetes/pkg/kubelet/checkpointmanager"
cperrors "k8s.io/kubernetes/pkg/kubelet/checkpointmanager/errors"
cdiapi "tags.cncf.io/container-device-interface/pkg/cdi"

configapi "github.com/NVIDIA/k8s-dra-driver-gpu/api/nvidia.com/resource/v1beta1"
Expand Down Expand Up @@ -181,6 +184,20 @@ func (s *DeviceState) Unprepare(ctx context.Context, claimRef kubeletplugin.Name
// Rely on local checkpoint state for ability to clean up.
checkpoint := newCheckpoint()
if err := s.checkpointManager.GetCheckpoint(DriverPluginCheckpointFileBasename, checkpoint); err != nil {
// Some errors returned by `GetCheckpoint` are permanent. Some of them
// may be retryable. For now, handle an explicit set of errors as
// permanent, and treat every other error as retryable.

if errors.Is(err, &json.UnmarshalTypeError{}) {
Copy link
Collaborator Author

@jgehrcke jgehrcke May 21, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Note: there is no overarching error type representing any error during Unmarshal().

As of https://pkg.go.dev/encoding/json#UnmarshalTypeError, Unmarshal() can actually return three different types of error:

Unmarshal returns an InvalidUnmarshalError.

Unmarshal returns a SyntaxError.

Unmarshal returns an UnmarshalTypeError

SyntaxError means "bad JSON", we also want to treat that as permanent.

What would be a concise way to test "is this error type any of these...?"

// May for example happen when a different version of this program
// wrote the JSON doc (using a different JSON schema).
return permanentError{fmt.Errorf("get checkpoint: unexpected schema (treat permanent): %w", err)}
}

if errors.Is(err, &cperrors.CorruptCheckpointError{}) {
return permanentError{fmt.Errorf("get checkpoint: bad checksum (treat permanent): %w", err)}
}

return fmt.Errorf("unable to get checkpoint: %w", err)
}

Expand Down
Loading