VMware HA Stretched Cluster

HA Stretched Cluster - the HA cluster "stretched" between two platforms on the VMware vSphere platform, using failsafe architecture of Virtual SAN.

The technology is created in 2011 as the "stretched" cluster of VMware vSphere 5 on the basis of technology and the equipment EMC VPLEX Metro with support of opportunities of VMware of HA and vMotion for fault tolerance and load distribution between geographically distributed by DPC^[1].

With an exit of VMware vSphere 5 there was a vSphere Metro Storage Cluster (vMSC) technology maintained for October 24, 2011 only for the solution EMC VPLEX, but maintained completely according to HCL regarding HA and vMotion technologies:

Representation of the "stretched" cluster, 2011

The component in the middle is the VPLEX Witness virtual machine which is the "witness" watching both platforms (it is located on the third platform - in one of two DPCs that accident did not affect), which can distinguish falling of communication on SAN and LAN network between DPC from falling of one of DPC (accident) due to monitoring of platforms on IP connection.

If the main website A fails, the mechanism of VMware of HA reboots its VM on the website B, servicing their input-output from the platform where there is a survived copy of virtual storage. The same happens at mass failure ESXi host servers on the main platform - virtual machines are restarted on hosts of the stretched website B cluster.

The situation with failures on the party of the website B where there are resistive loads is similar - its machines will pass to the website A. After recovery of the website (in both cases with failure both for A, and for B) - the virtual volume will be synchronized on both platforms (i.e. Failback is supported).

If only the network of management for ESXi hosts on one platform refuses, then VMware of HA will leave its virtual machines started as there is a mechanism for exchange of signals (heartbeats) through Datastore.

VMware vMotion, DRS and Storage vMotion - are also supported when using the solution EMC VPLEX Metro that allows to transfer loadings of virtual machines (both computing, and storage - vmdk-disks) between DPC without idle time of services. It opens opportunities not only for disaster tolerance, but also for such strategy as follow the sun and follow the moon (especially for them the EMC VPLEX Geo technology - 2000 km and 50 ms of latency is implemented).

Notes