APNIC's third morning: a maintenance window missing from the index

APNIC Put Two Mornings on Its Noticeboard. The Status Log Has a Third.

Why it matters

A network operator reading APNIC's service-update index would schedule around 3 and 5 November. The status record places another Brisbane maintenance window between them, with twenty named components and a different five-minute warning. The awkward part is not an outage claim. It is that the two public calendars disagree about what needed a place on the operator's calendar.

Theo March4 November 2026Reading Time 2 minLast update: 4 November 2026

APNIC published two adjacent migration notices; its status log records a distinct security-maintenance window between them.

Summary

APNIC's service-update index lists Brisbane network-migration notices for 3 and 5 November, each 07:00–09:00 UTC+10 and each warning of possible interruptions up to 15 minutes. The index has no listed 4 November notice at this cutoff.
APNIC's own status API records a separate 4 November 07:00–09:00 UTC+10 network-device security-maintenance event, twenty named components, a warning of possible interruptions up to five minutes, and a completion entry at 09:00. It gives no measured service interruption.

The morning between the notices

The 3 November announcement and 5 November announcement look like a tidy pair. Both reserve 07:00–09:00 in Brisbane, describe network migration in the data centre, and list the member portal, registry API, website, Academy, Blog, Orbit, DASH, RSC verifier and REx. Both warn that listed services may be interrupted for up to fifteen minutes. One window has passed; the other is still scheduled. Neither announcement reports a post-window impact measurement.

The public service-update index places those two dates on the noticeboard. It has no 4 November entry. APNIC's raw status record does: from 21:00 to 23:00 UTC on 3 November, which is 07:00–09:00 on 4 November in Brisbane. Its description is critical security updates to network devices, with possible interruptions up to five minutes. At the start it says In Progress; at the end, Completed. These are separate descriptions, so calling all three windows one migration would edit the record.

The middle entry names twenty components, not just a generic website. It reaches Whois Oceania and FTP, the MyAPNIC resource-administration surface, RPKI rsync, RRDP, TAL, repository, AS0 and publication paths, plus the registry API, RSC verifier, REx and web properties. That is a much more specific map of where an operator might have expected a maintenance state. The two website notices group their own affected services into nine names; the lists are different units and cannot be treated as a ten-component expansion.

A status label proves that APNIC marked the maintenance completed at 09:00 Brisbane time. It does not measure five minutes of lost access. The five- and fifteen-minute figures are prospective ceilings in different records, not observed downtime. Per-component interruption measurements would change that judgment.