Orchestration

Orchestration is the control-plane function of the Management Server: it schedules replication, sequences recovery, executes hooks, and keeps every action observable as a job.

What gets orchestrated

• Replication iterations — triggered per protection plan on the configured start time and interval, with per-VM scheduling (or synchronized sync times when Synchronize All VM Replication is enabled). Stopping, starting, and prioritization (AWS targets) are user-controllable.
• Recovery operations — test recovery, full recovery, migration, and reverse replication, each running as monitored recovery jobs with per-step progress.
• Point-in-time checkpoint jobs — created by the plan's checkpoint schedule, with retention-driven expiry, preservation, and deletion.
• Cleanup jobs — removal of test-recovered and recovered copies from the target infrastructure.

Recovery sequencing

Recovery orchestration is driven by the protection plan's configuration:

1. Plan-level pre-script runs first (when configured and selected).
2. VMs recover in the configured boot order — lower numbers first — with the configured boot delay in seconds between them, so dependencies (databases before app tiers) come up correctly.
3. Per-VM pre/post scripts wrap each VM's recovery; the post-script receives a JSON document with the recovered VM's IDs, IPs, and credentials.
4. Windows validation runs for Windows guests; Prep Nodes are powered on only for this work.
5. Plan-level post-script completes the operation.

Script runtimes are Shell, Bash, Python 3.x, and Golang; a failing script marks the job Partially Completed. The full execution matrix is in Create a plan.

API-aware parallelism

Recovery and replication run in parallel while respecting cloud API limits, quota constraints, and storage concurrency limits — using retry-aware workflows, backpressure-aware replication, and adaptive concurrency management. AWS's 100-parallel-snapshot-write cap, for example, is handled by replicating 80 disks and queueing the rest by configured priority. Plan quotas before scale events: Limits.

Observability

Every orchestrated action lands in Jobs (per plan, per VM, per disk), generates events, and raises alerts when attention is needed — with email delivery via notifications.

Related docs

• Scheduling

• Jobs

• Create a plan

• Limits