fact_check
Step 1Eligibility check
Every workload is scanned for migration eligibility first. OptOps only live-migrates what it can move safely — and tells you honestly when it can’t.
Container Live Migration
Move running pods. Drop nothing.
OptOps checkpoints eligible workloads with CRIU and restores them on another node — memory, process state, and on AWS even the pod IP and open TCP sessions come along.
deployed_codecheckout-svc
TCP Established
node-a · spot
Running · healthy
node-b · replacement
Restore target · standby
deployed_codepod · 10.0.4.17
Requests served
14,203
During migration
0 dropped
How a live migration works
photo_camera
Step 2Checkpoint
The running container is checkpointed with CRIU — memory, process tree, and open file descriptors captured in place. Databases get application-consistent safety handling as part of the checkpoint pipeline.
restart_alt
Step 3Restore
The pod resumes on the destination node exactly where it left off. On AWS, it keeps its pod IP — open TCP connections never notice the move.
Why it changes the economics
bolt
Spot interruption rescue
When the cloud reclaims a spot node, OptOps live-migrates the workload off it before the deadline — proven against real AWS FIS interruption drills, with rescue completing in as little as 33 seconds and the same pod IP on the other side.
balance
Disruption-free consolidation
Bin-packing normally means evicting pods and hoping they reschedule cleanly. With live migration, consolidation moves workloads instead of killing them — so the savings never cost you an incident.
database
Stateful workloads, handled honestly
Application-consistent safety engineering for Postgres, MongoDB, and MySQL is built into the migration pipeline. Workloads that can’t be moved safely are refused, not gambled with.
Platform support
We say exactly what works where. State preservation and connection preservation are different guarantees — here is the honest matrix.
| Platform | State-preserving migration | Connection preservation |
|---|---|---|
 Amazon EKS | Available | Available — same pod IP, unbroken TCP |
 Google GKE | Early access | On the roadmap |
 Azure AKS | Early access | On the roadmap |
Live migration applies to eligible workloads. GPU workloads and multi-container pods are not currently supported — eligibility is checked per workload before any move.
Watch a pod survive a node move
Book a demo and we'll live-migrate a running workload in front of you — counter ticking, connection open.
Book a Demo