Service status

State is derived from the probe agent on this host and from the two off-site checkers. A component is marked degraded when any single checker disagrees with the others.

Components

Component	State	30d availability	p95 answer	Checked from
dns-auth udp	operational	99.993%	2.1 ms	mow, ams, fra
dns-auth tcp	operational	99.991%	3.4 ms	mow, ams, fra
zone transfer	operational	100.000%	—	secondary set
probe agent	operational	99.87%	—	self
looking glass	limited	99.4%	—	self

The looking glass has been intentionally limited since 2026-06-02: public transcripts are no longer returned, only job acceptance. See the entry in the log below.

Incident log

Opened	Duration	Component	Summary
2026-06-02	—	looking glass	Change, not fault. Measurement transcripts withdrawn from public output after the endpoint was used to enumerate third-party hosts. Job intake stays open.
2026-05-14	41 s	dns-auth	Planned daemon restart to pick up a new signing policy. Secondaries answered throughout; a small share of clients saw one retry.
2026-04-27	6 m 12 s	probe agent	Upstream carrier maintenance dropped outbound ICMP. No effect on answers; RTT history for that window is missing and was not backfilled.
2026-03-09	2 m 03 s	dns-auth udp	Kernel conntrack table filled during a reflection attempt aimed at a neighbouring prefix. Rate limits were lowered afterwards; see peering notes.
2026-01-22	18 m	zone transfer	Provisioning database was read-only during a migration, so no new serials were generated. Serving copies were unaffected.

How to read these numbers

Availability is measured per protocol, not per host, because a host that answers TCP while dropping UDP is not usefully "up". The p95 figure is the answer latency observed by the checkers on the same continent and excludes network transit to the client.

Machine-readable form of this page is at /_api/status. It carries the same values and is the interface our own dashboards use; treat the HTML as a rendering of it.