So, in order to gain more experience with actually running IPv6 on the server- and client-side, we migrated all our users in the office to IPv6.
That sounds a lot harder than it actually is.
We installed radvd on one Linux machine connected to our office network. And that was basically it. radvd sends IPv6 announcements to all clients on the network. Windows, Mac, and Linux users automatically get IPv6 addresses that way. The one Windows XP machine you might have will need a manual “netsh int ipv6 install” once, and you are done.
So, we are using the following now:
- all our servers have native IPv6 connectivity
- our developers use SSH and HTTP via 6to4 from the office and home
- our servers talk to each other via IPv6 (mostly SSH and HTTP/HTTPS)
Because all the clients are dual-stacked, we have removed the IPv4 addresses from the main development systems’ DNS. So, where we had an A- and AAAA-record before, there is only one AAAA record left. This forces the dual-stacked system to connect via IPv6. For emergencies, there is always a hostname-4 entry left which points to the IPv4-address.
We had no problems whatsoever with any of the services we use internally so far (e.g. Apache, nginx, OpenSSH, MySQL, munin, exim, stunnel, etc.). The only exception is nullmailer (a simple SMTP injecter) whose IPv6 support gets broken in Debian every couple of months apparently.
Feel like sharing your IPv6 experience?