Evaluate migration from go-dockerclient to official Docker SDK

[CybotTM]

Related to mcuadros#228, mcuadros#47, mcuadros#216, mcuadros#215, mcuadros#214, mcuadros/ofelia PR mcuadros#386, PR mcuadros#396, PR mcuadros#397 (Docker SDK issues, CVEs, and memory leaks in go-dockerclient)

Summary

Consider migrating from github.com/fsouza/go-dockerclient to the official Docker SDK (github.com/docker/docker/client) to avoid ongoing compatibility issues and get better support.

Background

We've encountered several issues with go-dockerclient:

1. Event Listener Goroutine Leak (Issue #911)

• Issue: RemoveEventListener cause connection leak fsouza/go-dockerclient#911
• Status: Open since April 2022
• Impact: RemoveEventListener doesn't properly stop internal goroutines, causing "send on closed channel" panics and connection leaks
• Our workaround: Added panic recovery and cleanup delays

2. Mock Server Missing Events API (Issue #1157)

• Issue: testing.NewServer doesn't implement /events endpoint fsouza/go-dockerclient#1157 (just filed)
• Impact: testing.NewServer doesn't implement /events endpoint, causing tests to hang
• Our workaround: Added polling fallback alongside event listening

3. Maintainer Recommendation

From the go-dockerclient README:

This package was created before Docker had an official Go SDK and is still maintained.
For new projects, using the official SDK is probably more appropriate as go-dockerclient lags behind the official SDK.

Official Docker SDK Benefits

• Active maintenance by Docker team
• Feature parity with Docker API
• Better documentation
• No goroutine leak issues in event handling
• Official support for new Docker features

Migration Scope

Files using go-dockerclient

core/docker_operations.go
core/optimized_docker_client.go
core/container_monitor.go
core/runjob.go
core/execjob.go
core/localjob.go
core/servicejob.go
cli/daemon.go

API Surface Used

• Container lifecycle: Create, Start, Stop, Remove, Inspect
• Container logs: Logs
• Event listening: AddEventListener, RemoveEventListener
• Image operations: PullImage, InspectImage
• Network operations: CreateNetwork, ConnectNetwork
• Service operations (Swarm): CreateService, InspectService

Considerations

Pros

• Better maintained
• No known goroutine leaks
• Official Docker support
• Future-proof

Cons

• Migration effort required
• Need to update all Docker-related tests
• Potential API differences to handle

Acceptance Criteria

• Evaluate official SDK API compatibility
• Create migration plan with phased approach
• Migrate core Docker operations
• Update container monitoring to use official events API
• Update tests
• Remove go-dockerclient dependency
• Performance validation

References

• Official Docker SDK: https://pkg.go.dev/github.com/docker/docker/client
• go-dockerclient issue #911: RemoveEventListener cause connection leak fsouza/go-dockerclient#911
• go-dockerclient issue #1157: testing.NewServer doesn't implement /events endpoint fsouza/go-dockerclient#1157

[CybotTM]

✅ done with #270