You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: articles/azure-vmware/deploy-vsan-stretched-clusters.md
+16-11Lines changed: 16 additions & 11 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,13 +1,13 @@
1
1
---
2
-
title: Deploy vSAN stretched clusters
2
+
title: Deploy vSAN stretched clusters (Preview)
3
3
description: Learn how to deploy vSAN stretched clusters.
4
4
ms.topic: how-to
5
5
ms.service: azure-vmware
6
6
ms.date: 09/02/2022
7
7
ms.custom: references_regions
8
8
---
9
9
10
-
# Deploy vSAN stretched clusters
10
+
# Deploy vSAN stretched clusters (Preview)
11
11
12
12
In this article, you'll learn how to implement a vSAN stretched cluster for an Azure VMware Solution private cloud.
13
13
@@ -27,7 +27,7 @@ Stretched clusters allow the configuration of vSAN Fault Domains across two AZs
27
27
28
28
To protect against split-brain scenarios and help measure site health, a managed vSAN Witness is created in a third AZ. With a copy of the data in each AZ, vSphere HA attempts to recover from any failure using a simple restart of the virtual machine.
29
29
30
-
**vSAN stretched cluster**
30
+
The following diagram depicts a vSAN cluster stretched across two AZs.
31
31
32
32
:::image type="content" source="media/stretch-clusters/diagram-1-vsan-witness-third-availability-zone.png" alt-text="Diagram shows a managed vSAN stretched cluster created in a third Availability Zone with the data being copied to all three of them.":::
33
33
@@ -36,19 +36,24 @@ In summary, stretched clusters simplify protection needs by providing the same t
36
36
It's important to understand that stretched cluster private clouds only offer an extra layer of resiliency, and they don't address all failure scenarios. For example, stretched cluster private clouds:
37
37
- Don't protect against region-level failures within Azure or data loss scenarios caused by application issues or poorly planned storage policies.
38
38
- Provides protection against a single zone failure but aren't designed to provide protection against double or progressive failures. For example:
39
-
- Despite various layers of redundancy built into the fabric, if an inter-AZ failure results in the partitioning of the secondary site, vSphere HA starts powering off the workload VMs on the secondary site. The following diagram shows the secondary site partitioning scenario.
39
+
- Despite various layers of redundancy built into the fabric, if an inter-AZ failure results in the partitioning of the secondary site, vSphere HA starts powering off the workload VMs on the secondary site.
40
+
41
+
The following diagram shows the secondary site partitioning scenario.
40
42
41
43
:::image type="content" source="media/stretch-clusters/diagram-2-secondary-site-power-off-workload.png" alt-text="Diagram shows vSphere high availability powering off the workload virtual machines on the secondary site.":::
42
44
43
-
- If the secondary site partitioning progressed into the failure of the primary site instead, or resulted in a complete partitioning, vSphere HA would attempt to restart the workload VMs on the secondary site. If vSphere HA attempted to restart the workload VMs on the secondary site, it would put the workload VMs in an unsteady state. The following diagram shows the preferred site failure or complete partitioning scenario.
45
+
- If the secondary site partitioning progressed into the failure of the primary site instead, or resulted in a complete partitioning, vSphere HA would attempt to restart the workload VMs on the secondary site. If vSphere HA attempted to restart the workload VMs on the secondary site, it would put the workload VMs in an unsteady state.
46
+
47
+
48
+
The following diagram shows the preferred site failure or complete partitioning scenario.
44
49
45
50
:::image type="content" source="media/stretch-clusters/diagram-3-restart-workload-secondary-site.png" alt-text="Diagram shows vSphere high availability trying to restart the workload virtual machines on the secondary site when preferred site failure or complete partitioning occurs.":::
46
51
47
-
It should be noted that these types of failures, although rare, fall outside the scope of the protection offered by a stretched cluster private cloud. Because of this, a stretched cluster solution should be regarded as a multi-AZ high availability solution reliant upon vSphere HA. It's important you understand that a stretched cluster solution isn't meant to replace a comprehensive multi-region Disaster Recovery strategy that can be employed to ensure application availability. The reason is because a Disaster Recovery solution typically has separate management and control planes in separate Azure regions. Azure VMware Solution stretched clusters have a single management and control plane stretched across two availability zones within the same Azure region. For example, one vCenter Server, one NSX-T Manager cluster, one NSX-T Data Center Edge VM pair.
52
+
It should be noted that these types of failures, although rare, fall outside the scope of the protection offered by a stretched cluster private cloud. Because of those types of rare failures, a stretched cluster solution should be regarded as a multi-AZ high availability solution reliant upon vSphere HA. It's important you understand that a stretched cluster solution isn't meant to replace a comprehensive multi-region Disaster Recovery strategy that can be employed to ensure application availability. The reason is because a Disaster Recovery solution typically has separate management and control planes in separate Azure regions. Azure VMware Solution stretched clusters have a single management and control plane stretched across two availability zones within the same Azure region. For example, one vCenter Server, one NSX-T Manager cluster, one NSX-T Data Center Edge VM pair.
48
53
49
54
## Deploy a stretched cluster private cloud
50
55
51
-
Currently, Azure VMware Solution stretched clusters is in a limited availability phase. In the limited availability phase, you must contact Microsoft to request and qualify for support.
56
+
Currently, Azure VMware Solution stretched clusters is in the (preview) phase. While in the (preview) phase, you must contact Microsoft to request and qualify for support.
52
57
53
58
## Prerequisites
54
59
@@ -120,7 +125,7 @@ Azure VMware Solution stretched clusters are available in the following regions:
120
125
121
126
As of now, the only 3 regions listed above are planned for support of stretched clusters.
122
127
123
-
### What kind of SLA does Azure VMware Solution provide with the stretched clusters limited availability release?
128
+
### What kind of SLA does Azure VMware Solution provide with the stretched clusters (preview) release?
124
129
125
130
A private cloud created with a vSAN stretched cluster is designed to offer a 99.99% infrastructure availability commitment when the following conditions exist:
126
131
- A minimum of 6 nodes are deployed in the cluster (3 in each availability zone)
@@ -137,12 +142,12 @@ No. A stretched cluster is created between two availability zones, while the thi
137
142
- Scale out and scale-in of stretched clusters can only happen in pairs. A minimum of 6 nodes and a maximum of 16 nodes are supported in a stretched cluster environment.
138
143
- Customer workload VMs are restarted with a medium vSphere HA priority. Management VMs have the highest restart priority.
139
144
- The solution relies on vSphere HA and vSAN for restarts and replication. Recovery time objective (RTO) is determined by the amount of time it takes vSphere HA to restart a VM on the surviving AZ after the failure of a single AZ.
140
-
- Preview features for standard private cloud environments aren't supported in a stretched cluster environment. For example, external storage options like disk pools and Azure NetApp Files (ANF), Customer Management Keys, Public IP via NSX-T Data Center Edge, and others.
145
+
- Preview and recent GA features for standard private cloud environments aren't supported in a stretched cluster environment.
141
146
- Disaster recovery addons like, VMware SRM, Zerto, and JetStream are currently not supported in a stretched cluster environment.
142
147
143
148
### What kind of latencies should I expect between the availability zones (AZs)?
144
149
145
-
vSAN stretched clusters operate within a 5 minute round trip time (RTT) and 10 Gb/s or greater bandwidth between the AZs that host the workload VMs. The Azure VMware Solution stretched cluster deployment follows that guiding principle. Consider that information when deploying applications (with SFTT of dual site mirroring, which uses synchronous writes) that have stringent latency requirements.
150
+
vSAN stretched clusters operate within a 5-milliseconds round trip time (RTT) and 10 Gb/s or greater bandwidth between the AZs that host the workload VMs. The Azure VMware Solution stretched cluster deployment follows that guiding principle. Consider that information when deploying applications (with SFTT of dual site mirroring, which uses synchronous writes) that have stringent latency requirements.
146
151
147
152
### Can I mix stretched and standard clusters in my private cloud?
148
153
@@ -154,7 +159,7 @@ Customers will be charged based on the number of nodes deployed within the priva
154
159
155
160
### Will I be charged for the witness node and for inter-AZ traffic?
156
161
157
-
No. While in limited availability, customers won't see a charge for the witness node and the inter-AZ traffic. The witness node is entirely service managed, and Azure VMware Solution provides the required lifecycle management of the witness node. As the entire solution is service managed, the customer only needs to identify the appropriate SPBM policy to set for the workload virtual machines. The rest is managed by Microsoft.
162
+
No. While in (preview), customers won't see a charge for the witness node and the inter-AZ traffic. The witness node is entirely service managed, and Azure VMware Solution provides the required lifecycle management of the witness node. As the entire solution is service managed, the customer only needs to identify the appropriate SPBM policy to set for the workload virtual machines. The rest is managed by Microsoft.
0 commit comments