Deploying a vMSC across two sites using MacroSAN SDAS
search cancel

Deploying a vMSC across two sites using MacroSAN SDAS

book

Article ID: 329033

calendar_today

Updated On:

Products

VMware vSAN

Issue/Introduction

Introduction to MacroSAN SDAS

Symmetrical dual-active storage system (SDAS), which is simple and effective at “data protection and business continuity”, is newly developed by MacroSAN Technologies Co., Ltd. (MacroSAN). It supports the mirror-pairs based on enhanced features of Mirror, and both LUNs in a mirror-pair can provide the business simultaneously. When a disaster occurs in one data center, the business will continue running on the other data center without human intervention, to achieve that the value of RTO and RPO is 0.

MacroSAN SDAS, it not only supports close-distance dual-active storage system, but also support hundreds of kilometers or even more long-distance dual-active data center. You can adopt more efficient MESH mode in close-distance SDAS, and switch mode in long-distance SDAS, with the arbiter to improve network deployment. Arbiter is a software  which is deployed on a third-party server, when two LUNs in a mirror-pair cannot negotiate mirror role by themselves, they will use arbiter server to negotiate mirror roles.

Resolution

Technical requirements:

The following requirements must be fulfilled to implement a vMSC solution:
  • Use MacroSAN’s  storage devices such as MS2500G2, MS3000, MS3000G2, MS5000, MS5500, MS5500G2, MS7000, MS7000G2, MS9000G2, etc.
  • Each SP’s physical memory is larger than or equal to 8GB.
  • Software version is higher than or equal to ODSP_V1.2.27T04.
  • All SPs are interconnected by using SDAS’s high speed and low latency backend network, such as 10GE, 25GE, 40GE, 100GE, PCIe, etc.
  • SDAS’s backend network can use MESH mode or switch mode.
  • SDAS’s backend network delay RTT <= 5ms.
  • Require SDAS license on MacroSAN storage devices.
  • ESXi version: ESXi5.0/5.1/5.5/6.0/6.5/6.7.

Solution overview:



Technical suggestions and restrictions:
  • SDAS’s backend network is a private network, it cannot be shared with other services.
  • In order to reduce the fault area, the arbiter is not allowed to install on LUNs which are in the SDAS.
  • The arbiter network should be independent from SDAS’s backend network.

Tested Scenarios:
This table outlines the tested and supported failure scenarios when using a MacroSAN SDAS for VMware vSphere:
 

Scenario

MacroSAN

Procedure

Result (Observed VMware behavior)

Single-link failure of application server

Disconnect a link between nodeA in site A and application server.

 No impact.

Storage array failure in site A

Safe-remove disks which are belonging to  the array in site A.

No impact.

Storage array failure in site B

Safe-remove disks that are belonging to  the array in site B.

No impact.

One SP failure of nodeA in site A

Power off one SP of nodeA in site A.

No impact.

NodeA failure in site A

Cut off the power supply of nodeA in site A.

No impact.

NodeB failure in site B

Cut off the power supply of nodeB in site B.

No impact.

Multiple hosts failure in site A

Shut down multiple hosts in site A.

Virtual machines were halt by ESXi and restarted on normal hosts by vSphere HA.

Multiple hosts failure in site B

Shut down multiple hosts in site B.

Virtual machines were halt by ESXi and restarted on normal hosts by vSphere HA.

Failure of SDAS private links;
vSphere cluster   Management Network is working properly.

Disconnect all SDAS private links between nodes.

No impact.

The SDAS private link between nodes is normal.
Failure of vSphere cluster   management network

Disconnect management network of vSphere cluster  .

There is no virtual machine failover.

Failure of SDAS private links;
Failure of vSphere cluster  management network

Disconnect all SDAS private links between nodes; and disconnect management network of vSphere cluster  .

There is no virtual machine failover.

Failure of node management network

Disconnect management network of node.

No impact.

Failure of site A

Power off all devices in site A.

Virtual machines were halt by ESXi and restarted on normal hosts by vSphere HA.

Failure of site B

Power off all devices in site B.

Virtual machines were halt by ESXi and restarted on normal hosts by vSphere HA.

One SP failure of nodeA in site A, and storage array failure in site B

Power off one SP of nodeA in site A and safe-remove disks which are belonging to the array in site B.

No impact.

Failure of SDAS private links;
and link failure between arbiter server and site A.

Disconnect all SDAS private links between nodes; and disconnect the link between arbiter server and site A.

No impact.

Failure of SDAS private links;
failure of the link between arbiter server and site B.

Disconnect all SDAS private links between nodes; and disconnect the link between arbiter server and site B.

No impact.

Failure of arbiter server

Power off arbiter server.

No impact.



Additional Information