My users can't be followed form several instances

Hey there,

I’m facing a brand-new issue right now.
I’m hosting https://crazynoiisybizarre.town, and some users from several instances reported me that they can’t follow users from my instance. The opposite is possible tho.

The instance I know that can’t follow us : https://mastodon.social, https://mstdn.gnous.eu

I have searched the logs, but I can’t found a thing about that (or maybe I don’t know what to search).

Do you have any idea ?
Tagada ~

I’m in contact with @Madeorsk from https://aleph.land, who can’t follow @jusdepatate@crazynoisybizarre.town from his instance.

He gave me some log about that error : hastebin

As you can see line 19, there is an error about webfinger, but… I don’t know what to do next…
I hope this detail can be useful !

EDIT: After some testing, this bug affect all my users except one : @Dhveszak@crazynoisybizarre.town who has id=1.
I have recently set up another instance on the same server, and at start I forgot to set a different REDIS_NAMESPACE for both instances… Maybe this is related ?
How can I “clean” Redis ?

EDIT2: Something related to IPv6 seems more acurate.

Is this rather

crazynoisybizarre.town.

If yes, then you accept connections on IPv4.

You also have an IPv6 address:

crazynoisybizarre.town.	14286	IN	AAAA	2001:41d0:2:98b9:1:1:1:1

But this address cannot be reached:

> ping6 crazynoisybizarre.town
PING6(56=40+8+8 bytes) 2a01:4f8:a0:7383::9999:0 --> 2001:41d0:2:98b9:1:1:1:1
^C
--- crazynoisybizarre.town ping6 statistics ---
3 packets transmitted, 0 packets received, 100.0% packet loss
> nc -v -6 crazynoisybizarre.town 44343
nc: connect to crazynoisybizarre.town port 443 (tcp) failed: Operation timed out

Is this IPv6 address correct at all? I’d double check this - and make sure the machine has real IPv6 connectivity.

Also watch out for typos!