Azure VMSS Autoscaling

• Create a VMSS with image from compute gallery with autoscaling
  • min: 1
  • max: 5
  • scale out by 1 if cpu utilization > 60%
  • scale in by 1 if cpu utilization < 40%
    
    
    
    
• We have artificially stressed the VM in VMSS using

stress --cpu 8 --io 4 --vm 2 --vm-bytes 128M --timeout 100m -v

• The instances in VMSS before we started execution are 1, lets wait for 5-10 mins and check for the results
  
  
  
  
• By default we can scale using host based metrics
  • CPU
  • RAM
  • Disk utilization
• If you need to scale with Application level metrics we can use application insights Refer Here and also from diagnostic extension Refer Here

Rolling updates to VMSS

• Ensure your upgrade policy is not manual.
• Create a new version of the image in compute gallery/image definition and as the upgrade policy is automatic, it should upgrade the vmss with new image version.
  
• After the image rollout is complete you should be seeing the new image deployed in vmss without downtime.