Kubernetes and Checkpoint Restore
Talkat KubeCon + CloudNativeCon North America 2021 · on Thursday, 2021-10-14· at 15:25· Room 408 AB + Online·
summary
CRIU-based container checkpointing is shown working end-to-end: ptrace plus injected parasite code dumps process state and memory, file descriptors and FS deltas get written into a checkpoint archive, and restore recreates the exact pid/tree and resumes execution. Demos use Podman to export/import checkpoints (WildFly demo shows a restored counter continuing at 3), create fast clones (≈50% startup time saved), and migrate containers across hosts. A Kubernetes KEP (2008) for forensic checkpointing exists but needs CRI API changes to fully integrate.
AI-written summary · about
chapters (10)
- 0:00Start
- 1:15Definition
- 2:16Checkpoint/Restore in Userspace
- 3:07Container Live Migration
- 12:58Forensic Container Checkpointing
- 18:21Seize Process Using
- 19:50Parasite Code
- 22:07Checkpointing Finished
- 22:34clone () For Each PID/TID
- 22:51CRIU Morphs Itself
related talks
39mKubernetes and Checkpoint/Restore
32mForensic container checkpointing and analysis
39mContainer Live Migration
32mForensic Container Checkpointing
42mForensic container checkpointing and analysis - Adrian Reber
27mState of Checkpoint/Restore in Kubernetes – Now With GPUs
34mLetting Your Containers Dream of Electric Sheep
32mEnd-to-End Encryption for Container Checkpointing in Kubernetes