Zone replication of a VM - how to configure

Colm 26

hi,

about this article on Azure VM Zone redundancy

https://learn.microsoft.com/en-us/azure/virtual-machines/availability

Im curious on how to setup a VM for Zone redundancy with a load balancer?

Im unsure of how the load balancer fits into this?

eg. If a VM is running a service and has no external connections, could this just failover over to another zone and run ok?

does the VM that has failed over, get a different IP? is it in a different VNET?

Thanks

kobulloc-MSFT 23,496 Reputation points Microsoft Employee

2024-03-07T20:01:29.7733333+00:00

Hello, @Colm !

(https://docs.microsoft.com/en-us/answers/support/accepted-answers) to help increase visibility of this question for other members of the Microsoft Q&A community. If not, please let us know what is still needed in the comments so the question can be answered. Thank you for helping to improve Microsoft Q&A!

1 answer

TP 76,681 Reputation points

2024-03-05T16:07:02.9533333+00:00

Hi,

One way to make a workload more resilient to failure is to have it running across multiple Availability Zones. To do this you would need to run the workload on at least one VM in each of the zones (2-3 zones), AND the workload needs to have a way to synchronize state between the VMs OR the workload needs to be stateless.

For example, say you have a website that only consists of static files, so in effect it is stateless so long as each VM has current copy of static files. You would distribute the incoming requests across all VMs (this is where load balancer comes in). If one of the VMs failed, the load balancer would detect this and only send requests to the VMs in the other zones.

In the above example the client PCs are only communicating directly with the load balancer, so they are unaware of multiple VMs and the fact that they are in different zones.

Please let me know if that answers your question fully or if you need more details. Add a comment below.

Please click Accept Answer and upvote if the above was helpful.

Thanks.

-TP
Please sign in to rate this answer.

1 person found this answer helpful.
TP 76,681 Reputation points

2024-03-06T16:15:41.6633333+00:00

@Colm Any update? Was my answer clear, or do you still have questions?

Colm 26 Reputation points

2024-03-06T16:23:23.63+00:00

hi,

I know this makes sense for load too, but im talking about have 1 single VM with Zone availability enabled.

the status of the VM shows

"

Protect against hardware and software failures

"

the documentation lists that the VM will instantly come online in the event that the zone it is in fails but it needs a load balancer to make it work, i dont understand this need for the load balancer.

thanks

TP 76,681 Reputation points

2024-03-06T16:33:50.92+00:00

Which exact documentation page are you referring to? I can give you lots of examples of different ways to set up your VM to recover from various failures, but before I do that I want to try to address the specific page you have read.

Colm 26 Reputation points

2024-03-07T09:47:35.52+00:00

https://learn.microsoft.com/en-us/azure/virtual-machines/availability

"

Each Availability Zone has a distinct power source, network, and cooling. By designing your solutions to use replicated VMs in zones, you can protect your apps and data from the loss of a data center. If one zone is compromised, then replicated apps and data are instantly available in another zone.

"

there could be some confusion over what "replicated apps" implies that you've configured apps on multiple servers.

i think i actually found what i was looking for though

https://learn.microsoft.com/en-us/azure/site-recovery/azure-to-azure-how-to-enable-zone-to-zone-disaster-recovery

this definitely implies you need to configure like an ASR between the local Zones.

thanks

TP 76,681 Reputation points

2024-03-07T17:32:36.2533333+00:00

Thanks. Let me walk through some examples.

Example 1: Single VM. For this you could have the VM in say, Availability Zone 1, and use Azure Site Recovery to replicate to another zone or region. No load balancer involved in this example.

If the VM fails, it isn't "instantly" available in another zone, and there will likely be some data loss. There is a delay while you trigger failover, replacement VM and its required resources are created, VM starts up, etc. Crash-consistent recovery points (RP) are only created every 5 minutes, which is why I mention data loss. And this assumes the apps running inside of the VM are able to repair their data files from data in the latest RP.

So to summarize this example, think disaster recovery with small amount of downtime and some potential data loss.

Example 2: Multiple VMs with "intelligent" stateful workload. For this you could have VM in say, AZ 1, AZ 2, AZ 3. All VMs are running all the time, and handling requests. The app running in the VMs syncs data changes with the other VMs automatically. Real world example of this is Active Directory Domain Controllers. No load balancer involved in this example.

If one of the VM fails, the others are still working fine and have copies of the data, so only minor interruption because client PCs are aware of all three VMs and will automatically contact a different one if the first one doesn't respond. It is fair to think of this as "instant".

Example 3: Multiple VMs behind load balancer, stateful workload. This is similar to what I explained in my original answer, only stateful. You would have multiple VMs, each in different AZ. A zone-redundant load balancer sits in front of the VMs, and this is what client PCs communicate with. The app running on each VM stores its data in a backend that is resilient to zone failures such as Azure SQL Database with appropriate config.

If one of the VM fails, the load balancer detects this and stops sending new requests to that VM. The client PCs that were working with the failed VM may experience an error, but they can refresh their page/re-sign in/similar and continue working. It is fair to think of this as "instant".

Hopefully the above examples paint a better picture of some of the available options. Unfortunately there is no "easy button" for high availability that you can use for all apps/workloads.

TP 76,681 Reputation points

2024-03-11T13:30:57.4533333+00:00

@Colm Any update? Please click Accept Answer to close up this thread.

Thank you.
Sign in to comment