Following the failure of one of our controller nodes this evening we have had to perform emergency maintenance on a core segment of the network to reboot one of the central data switches.
The switch was refusing to learn the new data node addresses causing an imbalance in the network and affecting performance on the backbone.
After attempts to clear the issue manually it was decided that a reboot would be the quickest repair taking a brief 10 minute outage as the switch reloaded.
Apologies to affected cluster users. This affected 1 segment in our architecture, and did not affect the whole service. Customers with mirrored data nodes spread across the 3 cluster services would have noticed little or no disruption.
Normal service has been resumed in the time taken to write this notification.