SimpliVity Availability Zone "What If"

[J-Philippe]

Hi the community,

I wonder the follwing regarding a Stretched Cluster of 8 nodes and in fact Availability Zone.

What if :

Shuting Down (for maintenance for instance) in the rules of art 1 OVC (So at the cluster level 2 OVC) in each zone  ?

Does the IP failover works and no DU happen because the 2 hosts remain up and running and the other OVC have access to the storage of the concerning nodes ? No HA should be trigger and VM continue to run on that nodes ?

What if : 

Loosing 1 node in each zone ? My opinion is that DU event can happen for some VM that will have P and S data on both nodes.

Thank you for your help.

Jean-Philippe

[gustenar]

If you have Zones A & B for example, a stretch cluster will have a copy of that VM on each availability zone. If Zone A goes down, the VM will restart on zone B. If you face a failure on both zones at the same time, yes it is possible you will experience DU on some VMs. 

[Mahesh202]

Hi J-Philippe

In an HPE SimpliVity Stretched Cluster (also known as an Availability Zone), multiple nodes are distributed across different physical locations or zones to provide high availability and disaster recovery capabilities.

Scenario 1: Shutting down one OVC (2 nodes) in each zone for maintenance.

If you shut down one OVC in each zone, the VMs running on those nodes will experience a momentary interruption when the primary OVC goes down. However, since the HPE SimpliVity architecture is designed for high availability, the VMs will be automatically failed over to the secondary OVC (located in the same zone) with minimal downtime. This process is transparent to the VMs and should not trigger any HA events.

The remaining OVCs in each zone will still have access to the storage of the nodes within their respective zones, ensuring that VMs continue to run without any disruption. Once the maintenance is completed, powering on the OVCs will bring them back into the cluster, and VMs can be redistributed back to their original locations if needed.

Scenario 2: Losing one node in each zone.

If you lose one node in each zone, the HPE SimpliVity Stretched Cluster is designed to handle such scenarios without any data loss. Each VM in the cluster has primary (P) and secondary (S) data copies across different nodes to ensure redundancy and protection.

When a node goes down, VMs with primary and secondary data copies on that node are automatically failed over to their secondary location, which is on another node within the same zone. This process is again transparent to the VMs and should not trigger a HA event.

While the VMs are running on their secondary data copies, there may be a slight performance impact due to increased I/O activity on the nodes hosting the secondary copies. However, this impact should be temporary and should not cause any data unavailability or data loss.

Overall, the HPE SimpliVity Stretched Cluster is designed to provide continuous availability and data protection, even in the face of node or OVC failures, ensuring that your VMs can run without interruption and your data remains protected. It is important to follow best practices and ensure that the cluster is properly configured to handle these scenarios effectively.

Hope this helps.!!

Regards
Mahesh.

[J-Philippe]

Hi Mahesh, thank you. 

I think your wrong on scenario 2 because in a Stretched cluster the availabilty zone ensure that P and S data are not within a same zone.

So if I lost 1 node in each zone I may face a situation where the VM "TEST" may have P data on the failed node in zone 1 and S data on the second failed node in zone 2.

Regarding the scenario 1 if it really works it is a good news as the SimpliVity support team for Europe don't think the same. I'm going to look deeper to ensure it is possible.

Many thanks

[gustenar]

@J-Phillippe,

Check this video that explains both scenarios: 

https://support.hpe.com/hpesc/public/videoDisplay?videoId=vtc00000092en_us

[J-Philippe]

@gustenar 

Hi Gunstar thanks for the link that's exactly what I'm saying / writing  4min19 sec of the video Declan explain that the concerning VM that have P and S on the concerning node will be unavailable. So losing 1 node in each zone at the same may impact some VM availabilty but sure not all the VM of the cluster.

My concern now is to find a technical doc that confirm what Mahesh is saying : that shuting down 1 OVC in each zone is not a pb as the node (then the underlying storage) are still available and 1 IP failover can happen in both zone to manage access to the concerning nodes.

Thanks a lot.