Hacker News Reader

Comment by niftich | Hacker News Reader

Fijivillage.com reports: "All websites and apps hosted in Fiji with the dotcom.fj suffix are currently down and this has also affected Vodafone’s M-PAiSA services.

This is due to an outage in the University of the South Pacific hosted dotcom.fj domain." [1][2]

[1] https://www.fijivillage.com/news/All-websites--apps-in-Fiji-...

[2] https://twitter.com/fijivillage/status/1501070675691278339

View on HN

And this folks is one of the reasons you don't use a 1 minute TTL on your DNS responses...

shamefully opens domain registrar and changes TTL

15 min?

I would recommend an hour for almost everything except where very fast updates is expected, in which case 5m is my lowest number (I work at a registrar).

Obviously registrars allow 1m - is there a reason not too besides extra load on DNS servers?

If the DNS server is out for more than 1 minute you'll get an outage. You also increase latency by preventing anybody from caching the response for more than 1 minute.

There are rumors of DNS resolvers deciding that some TTLs are “too low” to be valid, and applying their own default TTL value instead, thereby negating any benefit a low TTL would have had.

I use 24 hours on anything I'm not planning on making changes to. High TTL is a better experience for customers because they'll probably have it in local-ish caches even if there is internet routing disruption.

And 10 mins on anything I'm about to make changes to. That means if I accidentally make the wrong change, the 'blast radius' is minimized.

Obviously, when changing 24h down to 10 mins, keep a close eye on DNS server load, packet loss on links close to it, etc. If in doubt, raise and lower ttl's slowly.

Reply on news.ycombinator.com