Skip to content

Scale vJailbreak

vJailbreak can be scaled to perform multiple migrations in parallel by deploying additional agents, enabling greater efficiency and workload distribution.

Additional agents can be created in the Agents tab of the vJailbreak dashboard using the “Scale Up” button. You will need to choose the destination OpenStack credentials, the size of the agent VM(s), and the number of agent nodes up to a maximum of 5 per scale up. Additional agent nodes can be scaled up in batches of 5, providing the flexibility to change agent VM sizes to help with throttling network traffic.

Agent Node Sizing and Migration Capacity

Each migration running on an agent node consumes the following resources:

ResourceRequestLimit
CPU1 core2 cores
Memory1 GiB3 GiB
Ephemeral Storage3 GiB3 GiB

Calculating Concurrent Migrations per Agent

The number of concurrent migrations an agent node can handle depends on its available resources. While Kubernetes uses resource requests for scheduling decisions, the actual resource consumption during migration is closer to the limits. Therefore, consider both when planning capacity:

Scheduling Capacity (based on requests):

  • Maximum Concurrent Migrations = min(Available CPU / 1 core, Available Memory / 1 GiB, Available Storage / 3 GiB)

Actual Runtime Capacity (based on limits):

  • Maximum Concurrent Migrations = min(Available CPU / 2 cores, Available Memory / 3 GiB, Available Storage / 3 GiB)

For safe capacity planning, use the limits-based calculation to ensure migrations have sufficient resources during peak usage.

Below are recommended OpenStack flavors for agent nodes based on desired migration capacity. Reserve approximately 20-25% of resources for system overhead (OS, K3s, monitoring, etc.):

Agent FlavorvCPUsRAMStorageConcurrent Migrations (per agent)Use Case
Small816 GiB60 GiB2-3Small-scale migrations, testing
Medium1632 GiB100 GiB5-7Standard production workloads
Large3264 GiB200 GiB10-14High-throughput migrations
X-Large4896 GiB300 GiB15-21Maximum parallel migrations

Example Calculation for Medium Flavor (16 vCPU, 32 GiB RAM):

  • Available CPU after overhead: ~12 cores → 12 / 2 = 6 migrations
  • Available Memory after overhead: ~24 GiB → 24 / 3 = 8 migrations
  • Effective capacity: 6 concurrent migrations (limited by CPU)

Best Practices

  • Network bandwidth is often the bottleneck. Monitor network utilization and adjust agent count/size accordingly.
  • Storage I/O on the agent node should be sufficient for temporary disk operations during migration.
  • Start with Medium flavors and scale up based on observed resource utilization and network capacity.
  • Distribute migrations across multiple smaller agents rather than one large agent for better fault tolerance.
  • Monitor agent resource usage via the vJailbreak dashboard or Prometheus metrics to optimize sizing.

Agent nodes can be scaled down by selecting the agent and using the “Scale Down” button.

Logging into Agent VMs

Agent VMs use the same login process as the primary vJailbreak VM:

  • Username: ubuntu
  • Default Password: password
  • On first login, you will be prompted to change the password immediately.

Each agent must also have a copy of the VMware VDDK libraries in their /home/ubuntu directories.

  • Copy the latest version of the VDDK libraries for Linux into /home/ubuntu of the new agents. Untar it to a folder name vmware-vix-disklib-distrib in /home/ubuntu directory.