Cram sheet — Create and configure virtual machines

The heaviest topic in the unit. Two of its numbers are fixed at creation and can never be changed, and both are examined: an availability set's domain counts, and a scale set's orchestration mode.

Availability sets: the numbers, and what they buy

• Each availability set can have up to 3 fault domains and 20 update domains, and these settings cannot be changed after the set is created.
• Two or more VMs in an availability set is what meets the 99.95% SLA. One VM in a set buys nothing.
• There is no extra cost for the availability set itself — you pay only for the VM instances.
• During planned maintenance, only one update domain is restarted at a time, and a restarted update domain gets 30 minutes to recover before the next one is touched.
• With five update domains and six VMs, the sixth lands in the same update domain as the first. The distribution wraps.
• Availability sets give lower VM-to-VM latency than zones, because the machines sit closer together — and that is exactly why they do not survive a datacenter-level failure.

Resizing is disruptive, and sometimes collective

• Deallocation may be required if the new size is not available on the hardware cluster currently hosting the VM. If a size is missing from the portal list, stopping the VM may reveal more.
• Even when deallocation is not required, changing the size of a running VM restarts it. Treat every resize as disruptive.
• ⚠ Deallocating releases any dynamic IP addresses. The OS and data disks are not affected. That split — addresses lost, disks kept — is the examinable half.
• ⚠ Inside an availability set it gets worse: if the new size is not available on the current cluster, you must deallocate every VM in the set to resize one of them.

Disks and scale sets

• Five disk types: Ultra Disk, Premium SSD v2, Premium SSD, Standard SSD, Standard HDD.
• ⚠ Ultra Disk and Premium SSD v2 cannot be used as an OS disk. The two fastest options are data disks only, which quietly eliminates them from any stem about boot performance.
• Scale sets have two orchestration modes, and the orchestration mode is defined when you create the scale set and cannot be changed or updated later. Flexible spreads VMs across fault domains and offers high-availability guarantees for up to 1,000 VMs.

Traps

Check yourself

Where these figures come from

Every figure above was read from the raw documentation below on the day this sheet was written. The sha1 is git hash-object over the bytes as fetched, so a doc that changes underneath this sheet can be detected rather than assumed.