::: Virtual Aleph ::: Virtualization Techniques: #VMworld day 3 part 2: Architecting and Operating a VMware vSphere Metro Storage Cluster Privacy Policy

11 October, 2012

#VMworld day 3 part 2: Architecting and Operating a VMware vSphere Metro Storage Cluster

...after a quick round trip at Solutions Exchange, where I've spoken with Rick from Veeam, I've ran back to attend INF-BCO1159, with great speakers: Lee Dilworth and Duncan Epping.

Title of the session "Architecting and Operating a VMware vSphere Metro Storage Cluster".

Two possible architectures are feasible, but first look the VMware HCL for Metrocluster support.

The two architectures are based on Uniform or Non Uniform host based access to storage.

It's not so easy to set up a vSphere Metro cluster :

Things to keep in mind:
RTT latency must be under 10 milliseconds

- HA, DRS and SDRS are NOT site aware

- stretched VLANs are not loved by network guys

- site affinity must be taken into count ( locality matters!): site awareness could be obtained using DRS affinity

or via Storage DRS and Datastore Clusters; henve avoiding unnecessary site to site migrations.

- HA considerations:
isolation addresses must be specified at each site using advanced options.

Moreover each sites need a heartbeat Datastore defined to be feasible for both sites to update region for storage local to that site.

Permanent Device Loss (PDL) MUST be configured.

One HA master must be present on each site.


Some terminology:
APD, PDL and Split Brain

Full failure in one site could be solved with Metrocluster

SRM can be used to protect the production site: anyway additional considerations must be taken into count.