Users Guide

Metro HA with cross-connect
Metro node Metro HA with cross-connect (front end ports of metro node are cross-connected) can be deployed where the
metro node clusters are separated by 1 ms latency RTT or less. Metro node Metro HA combined with cross-connect eliminates
RTO for most of the failure scenarios.
Metro HA with cross-connect failure management
This section describes how metro node Metro HA with cross-connect rides through failures of hosts, storage arrays, clusters,
metro node Witness, and the inter-cluster link.
Host failure
If hosts at one site fail, then VMware HA restarts the virtual machines on the surviving hosts. Since surviving hosts are
connected to the same datastore, VMware can restart the virtual machines on any of the surviving hosts.
Cluster failure
If a metro node cluster fails:
● Metro node Witness guides the surviving cluster to continue.
● VMware re-routes I/O to the surviving cluster.
● No disruption to I/O.
Storage array failure
If one or more storage arrays at one site fail:
● All distributed volumes continue I/O to the surviving leg.
● No disruption to the metro node clusters or the virtual machines.
● I/O is disrupted only to local virtual volumes on the metro node cluster attached to the failed array.
Metro node Witness failure
If metro node Witness fails or becomes unreachable (link outage):
● Both metro node clusters call home to report that metro node Witness is not reachable.
● No disruption to I/O, metro node clusters, or the virtual machines.
Inter-cluster link failure
If the inter-cluster link fails:
● Metro node Witness guides the preferred cluster to continue.
● I/O suspends at the non-preferred cluster.
● VMware re-routes I/O to the continuing cluster.
● No disruption to I/O.
The following table summarizes how metro node HA with cross-connect manages failures.
Table 4. How metro node Metro HA recovers from failure
Failure description Failure handling
Host failure (Site 1) VMware HA software automatically restarts the affected
applications at Site 2.
Metro node cluster failure
(Site 1)
Metro node Witness detects the failure and enables all
volumes on the surviving cluster.
30 Integrity and resiliency