Customer Feature Request: Add "high-water" measurement for tcp-clients
Since we changed tcp-clients limit enforcement in response to CVE-2018-5743, several customers have contacted us asking how to set limits appropriately.
Yoshitaka Aharen of M-root has offered what we think would be a useful suggestion which would help him and other operators who want to scale limits appropriately. He asks if named can keep track of a "high-water" limit for the number of simultaneous TCP clients, correctly pointing out that this is likely to be of more help to customers than random periodic polling via 'rndc status' which, if not performed frequently, could easily miss spikes in usage.