Images, Azure Compute Gallery and VM scale sets

Load the network IDs and create the lab group:

$v = Get-Content C:\azure-lab\vnet-saat.json | ConvertFrom-Json
az group create --name rg-saat-l20 --tags course=saatvedha-azure lesson=20 -o none

1Build the source VM

Save the cloud-init file, then create build01 (about two minutes):

$y = @'
#cloud-config
write_files:
  - path: /var/www/index.html
    content: |
      <h1>img-web 1.0.0</h1>
  - path: /etc/systemd/system/saat-web.service
    content: |
      [Unit]
      Description=SaatVedha lab web page
      After=network-online.target

      [Service]
      ExecStart=/usr/bin/python3 -m http.server 80 --directory /var/www
      Restart=always

      [Install]
      WantedBy=multi-user.target
runcmd:
  - systemctl daemon-reload
  - systemctl enable --now saat-web
'@
[IO.File]::WriteAllText('C:\azure-lab\img-web-init.yaml', $y.Replace("`r`n", "`n"))
az vm create --resource-group rg-saat-l20 --name build01 --image Canonical:ubuntu-24_04-lts:server:latest --size Standard_B2ats_v2 --subnet $v.'snet-web' --public-ip-address '""' --nsg '""' --admin-username azureuser --generate-ssh-keys --custom-data C:\azure-lab\img-web-init.yaml --tags course=saatvedha-azure lesson=20 -o none

Check the page is served:

az vm run-command invoke --resource-group rg-saat-l20 --name build01 --command-id RunShellScript --scripts 'cloud-init status --wait; curl -s localhost' --query "value[0].message" -o tsv

You should see this:

Enable succeeded: 
[stdout]
status: done
<h1>img-web 1.0.0</h1>

[stderr]

2Generalize it

Run this:

az vm run-command invoke --resource-group rg-saat-l20 --name build01 --command-id RunShellScript --scripts 'waagent -deprovision+user -force' -o none
az vm deallocate --resource-group rg-saat-l20 --name build01
az vm generalize --resource-group rg-saat-l20 --name build01

Deprovision removes azureuser, SSH host keys and the hostname. If the first command reports a timeout, carry on: the deprovision still ran. build01 can never start again after this; it exists only to be copied.

3Create the gallery, image definition and version 1.0.0

Run this (the version takes 10 to 15 minutes):

az sig create --resource-group rg-saat-l20 --gallery-name galsaat --tags course=saatvedha-azure lesson=20 -o none
az sig image-definition create --resource-group rg-saat-l20 --gallery-name galsaat --gallery-image-definition img-web --publisher SaatVedha --offer web --sku ubuntu2404 --os-type Linux --os-state Generalized -o none
$vmId = az vm show --resource-group rg-saat-l20 --name build01 --query id -o tsv
az sig image-version create --resource-group rg-saat-l20 --gallery-name galsaat --gallery-image-definition img-web --gallery-image-version 1.0.0 --virtual-machine $vmId --replica-count 1 --storage-account-type Standard_LRS --tags course=saatvedha-azure lesson=20 -o none
az sig image-version list --resource-group rg-saat-l20 --gallery-name galsaat --gallery-image-definition img-web --query "[].{version:name, state:provisioningState, regions:join(', ', publishingProfile.targetRegions[].name)}" -o table
az vm delete --resource-group rg-saat-l20 --name build01 --yes

You should see this:

Version    State      Regions
---------  ---------  -------------
1.0.0      Succeeded  Central India

The CLI set the definition to Gen2 with Trusted Launch support, matching the Ubuntu 24.04 build VM.

4Create the scale set across two zones

Run this (about three minutes):

$img = az sig image-version show --resource-group rg-saat-l20 --gallery-name galsaat --gallery-image-definition img-web --gallery-image-version 1.0.0 --query id -o tsv
az vmss create --resource-group rg-saat-l20 --name vmss-web --image $img --vm-sku Standard_B2ats_v2 --instance-count 2 --zones 1 2 --orchestration-mode Uniform --upgrade-policy-mode Manual --subnet $v.'snet-web' --lb '""' --admin-username azureuser --generate-ssh-keys --tags course=saatvedha-azure lesson=20 -o none
az vmss list-instances --resource-group rg-saat-l20 --name vmss-web --query "[].{id:instanceId, name:name, zone:zones[0], latest:latestModelApplied}" -o table

You should see something like this:

Id    Name        Zone    Latest
----  ----------  ------  --------
0     vmss-web_0  1       True
2     vmss-web_2  2       True

The IDs may skip a number: that’s overprovisioning. One instance in each zone, both on the latest model. There’s no load balancer yet; it joins in Lesson 51.

5Check both copies serve the page

Run this:

$ids = az vmss list-instances --resource-group rg-saat-l20 --name vmss-web --query "[].instanceId" -o tsv
foreach ($id in $ids) { az vmss run-command invoke --resource-group rg-saat-l20 --name vmss-web --instance-id $id --command-id RunShellScript --scripts 'hostname; curl -s localhost' --query "value[0].message" -o tsv }

You should see something like this:

Enable succeeded: 
[stdout]
vmss-web000000
<h1>img-web 1.0.0</h1>

[stderr]
Enable succeeded: 
[stdout]
vmss-web000002
<h1>img-web 1.0.0</h1>

[stderr]

Different hostnames, same app, nothing installed by hand.

6Add autoscale rules

Run this:

az monitor autoscale create --resource-group rg-saat-l20 --resource vmss-web --resource-type Microsoft.Compute/virtualMachineScaleSets --name as-vmss-web --min-count 1 --max-count 2 --count 2 -o none
az monitor autoscale rule create --resource-group rg-saat-l20 --autoscale-name as-vmss-web --condition "Percentage CPU > 70 avg 5m" --scale out 1 -o none
az monitor autoscale rule create --resource-group rg-saat-l20 --autoscale-name as-vmss-web --condition "Percentage CPU < 25 avg 10m" --scale in 1 -o none
az monitor autoscale rule list --resource-group rg-saat-l20 --autoscale-name as-vmss-web --query "[].{metric:metricTrigger.metricName, op:metricTrigger.operator, threshold:metricTrigger.threshold, action:scaleAction.direction}" -o table

You should see this:

Metric          Op           Threshold    Action
--------------  -----------  -----------  --------
Percentage CPU  GreaterThan  70.0         Increase
Percentage CPU  LessThan     25.0         Decrease

Maximum 2 keeps the lab inside your quota. With idle CPU, autoscale may scale in to 1 after about 10 minutes; that’s the rule working. In production, the scale-in threshold sits well below scale-out, so the set doesn’t flap between sizes.

✅ Checkpoint: Image version 1.0.0 is in galsaat, vmss-web runs copies in zones 1 and 2, and both rules are listed.