One of our edge routers / BGP speakers started to run into difficulties at around 00:15 this morning. After a brief investigation it was deemed that the router (which had been up for over 300 days) was starting to exhibit “out of memory” behaviour which we believe can be attributed to a memory leak caused by excessive route flaps from one of our upstream peers
We attempted to correct the issue through software however after several unsuccessful attempts to close the affected daemon it was decided to reboot the affected node.
During the reboot customers connected through our DSL network may have noticed a short period of time while their connections bounced to one of our failover/backup connections.
The reboot commenced at 00:53 and was completed by 01:25
Once the reboot had completed, existing customer connections were migrated back through the primary feeds and normal service was restored