NIST warnt davor, dass mehrere seiner Internet-Zeitdienstserver aufgrund eines Stromausfalls ungenau sein könnten – Boulder-Server „verfügen nicht mehr über eine genaue Zeitreferenz“

My RPi with the GPS antenna is just fine.

/brag

* Power was disrupted Dec 17, 22:23 UTC

* Critical standby generator failure

* affected hosts: time-a-b.nist.gov through time-e-b.nist.gov, along with ntp-b.nist.gov, which is used for authenticated NTP.

* Time drifted by roughly four microseconds

* NIST has not provided a firm estimate for when full service will be restored at the Boulder campus.

Unclear if these hosts are still included in the time.nist.gov round-robin pool.

Also:

>The Boulder incident follows another Internet Time Service disruption on December 10 at NIST’s Gaithersburg, Maryland site, where an atomic time source failure caused a time step of approximately minus 10 milliseconds on affected hosts.

My atomic clock set itself 3 hours into the future this week. For no apparent reason.

For anyone that four microseconds is going to screw over they should already be using multiple sources. Curious to think about if taking these affected servers offline would be a better choice, or, tying them to another, slightly less reliable/accurate source like a GPS receiver, rather than just letting them drift with notice?

when power was down, the time was provided by: Casio.

Unless you are running scientific experiments, 4 µs is below the threshold your synchronization protocol (NTP) is capable of detecting. By like 2–3 orders of magnitude.

Ironically, the article does not have a date or time stamp