VM Scheduler

What It Does

The VM Start/Stop scheduler manages the Pexip Infinity VMs (Management Node and Conference Node) by starting them during service hours and deallocating them outside of service hours. This reduces Azure compute costs by only running VMs when the Verify by Video service is available to citizens.

Managed VMs

Environment	Management Node	Conference Node
Test	vm-node-mgmt-test	vm-node-conf-test
Production	vm-node-mgmt-core	vm-node-conf-core

Schedule

The scheduler runs 4 times per hour on a cron schedule:

55,35,25,5 * * * *

Each run determines the appropriate action based on service hours (sourced from the IAS /video/v2/service_hours endpoint):

• Within service hours + VMs deallocated → Start VMs, then run the post-start script (seeds the VMR pool)
• Outside service hours + VMs running → Deallocate VMs
• No action needed → Skip

Workflow Files

Workflow	Environment	Target Resource Groups
vm-start-stop-test.yml	test	rg-node-management-test, rg-node-conference-test
vm-start-stop-prod.yml	prod	rg-node-management, rg-node-conference

Pausing the Scheduler

During maintenance windows (e.g., Pexip upgrades, infrastructure changes), operators should pause the scheduler to prevent it from starting or stopping VMs unexpectedly.

Via the GitHub UI

1. Navigate to the repository → Actions tab
2. Select the workflow (e.g., “VM Start/Stop Scheduler (Production)”)
3. Click the ⋯ menu (top right of the workflow page)
4. Select Disable workflow

Via the GitHub CLI

# Disable the production scheduler
gh workflow disable "vm-start-stop-prod.yml" --repo Maximus-Canada/verify-by-video

# Disable the test scheduler
gh workflow disable "vm-start-stop-test.yml" --repo Maximus-Canada/verify-by-video

Resuming the Scheduler

Via the GitHub UI

1. Navigate to Actions → select the disabled workflow
2. Click Enable workflow (banner at top of workflow page)

Via the GitHub CLI

gh workflow enable "vm-start-stop-prod.yml" --repo Maximus-Canada/verify-by-video

When to Pause

• Pexip upgrades - VMs must remain on during the upgrade process and not be deallocated by the scheduler
• Infrastructure maintenance - any Azure operations that require VMs to be in a specific state
• Troubleshooting - when investigating VM issues and you need to prevent state changes

Important Notes

• Disabling a workflow prevents all triggers (schedule, manual dispatch, and any other). To run it one-off while disabled, re-enable it, run manually, then disable again.
• Disabled workflows retain their run history - nothing is lost.
• Anyone with write access to the repository can disable/enable workflows.
• Consider notifying the team when disabling a scheduler so others know VMs may be in an unexpected state.

Manual Override

Both scheduler workflows support workflow_dispatch with override inputs to force specific actions regardless of the schedule:

Input	Effect
runScaleDown: true	Force deallocate VMs (bypass schedule check)
runScaleUp: true	Force start VMs (bypass schedule check)
runPostStartScript: true	Force post-start script (bypass schedule check)

This is useful for manually starting VMs outside service hours (e.g., for evening maintenance) without needing to fully enable the scheduler.