stretched cluster testing

[Convergo]

What would be a good scenario to test a stretched cluster with four nodes, two nodes in an availability zone per datacenter ?

when running a node in maintenance node, I can't shutdown the virtual controller of the second node in the same availability zone.

Any creativity is being appreciated

[Sanika]

Hi Convergo,

I see you are asking for a suitable configuration for testing a four-node stretched cluster, with two nodes in an availability zone per datacenter.

Before that ensure you follow the below guidelines to use stretched clusters (from HPE SimpliVity Administration Guide):

1. Existing cluster size and network guidelines apply to Stretched Clusters.
2. Place the Arbiter in a third physical site.
3. Additional requirements for the Stretched Clusters feature are as follows:
(a) Stretched Clusters are supported with an even number of HPE OmniStack hosts. Hosts must be equally distributed in the two Availability Zones within the cluster. Stretched Clusters are not supported with an odd number of hosts.
(b) Stretched Cluster configurations support up to 16 HPE OmniStack hosts.
4. The recommended max latency value between availability zones is 2ms, a 10Gbps connection is recommended.

Here are some additional video links that could be of help:

HPE SimpliVity - Stretch Clusters - Overview
https://support.hpe.com/hpesc/public/videoDisplay?videoId=vtc00000092en_us

HPE SimpliVity - Stretch Clusters - Configure
https://support.hpe.com/hpesc/public/videoDisplay?videoId=vtc00000095en_us

HPE SimpliVity - Stretch Clusters - Unconfigure
https://support.hpe.com/hpesc/public/videoDisplay?videoId=vtc00000091en_us

To test a a four-node stretched cluster, with two nodes in an availability zone per datacenter, you may follow the below steps:

1. Ensure that the stretched cluster is functional with all 4 nodes (2 in each datacenter's Availability Zone) operational. The data of the VM should be distributed and accessible to all nodes.

2. Make sure all the hosts in one availability zone are powered off at once, then verify that VMs in another availability zone are still available.

Similarly, you may power off one host in each availability zone and ensure that VMs running on the remaining hosts are still available.

3. Place one host in each availability zone into maintenance mode and verify that the remaining hosts are still live.

4. Verify that a VM is available both during and after a live migration by moving it from one location to another.

5. Simulate a site failure by bringing down the network connection between the 2 datacenters.

I understand that you are not able to shutdown the OmniStack Virtual Controller (OVC) of the second node while the first node is put under maintenance mode.
This is because the vSAN may loose the quorum and the cluster will become unavailable. The cluster needs atleast two active hosts in each availability zone to maintain the quorum.

If you are still unable to shut down the OVC, you may contact HPE SimpliVity support for further assistance.

Hope this helps.

Regards,
Sanika

[Aditya_A]

Hi,

Please let me know if you have further queries.