• NICT said part of its redundant public time-service system distributed incorrect timestamps for 11 minutes during maintenance
  • The service was operating normally when NICT disclosed the incident, while the cause and size of the timing error remained under investigation

The fact

Japan’s National Institute of Information and Communications Technology said on 11 September that part of its public internet time service had distributed incorrect timestamps during maintenance two days earlier. The incident lasted from 10:20 to 10:31 Japan Standard Time. NICT said part of its redundant system was affected. The service was operating normally when the notice was published, and its other time-distribution services were unaffected. Engineers were still investigating the detailed cause.

Before the maintenance, NICT had warned that connections to some servers could be interrupted between 10:00 and noon on 9 September, with other servers expected to keep the service available. The later disclosure identified a different problem: part of the system had supplied incorrect time rather than simply becoming unavailable. The public service uses the Network Time Protocol, or NTP, to distribute time over the internet. NICT did not disclose how far the timestamps were wrong, how many users received them or whether connected systems adjusted their clocks.

The assessment

NICT’s notice is more awkward than a simple outage because the service did not merely disappear. Some servers kept responding, but with the wrong time. A normal availability check could therefore have stayed green while clients were receiving bad data.

What matters now is whether any customer systems actually trusted those responses. A client using several genuinely independent time sources may have rejected the bad reading when other sources disagreed. A system relying too heavily on one source could have behaved differently. The 11-minute window gives infrastructure teams a clear place to look: NTP client logs, clock-offset records and monitoring alerts should show whether an affected server was selected and whether the local clock moved.

For BTW readers, this is a reminder that resilience is not only about keeping a service online. For timing infrastructure, the service also has to be right. NICT has not reported widespread downstream failures, so the next step is to establish which systems saw the incorrect timestamps, which accepted them and whether monitoring caught the problem before it affected applications.

What to watch

Watch for NICT to disclose the cause, the size of the timing error and how the problem was detected. A specific prevention measure would show what changed after the incident and help users interpret their own records from the affected period.