Quick ask- what happens when you connect via the public network?

Fully working when connected through public network.

And, another thing is, since we switched to public network, the amount of bandwidth cost increased to 16576GB from 2827GB. However, the total amount of our websites serves only 6GB, was this a correct value of how it was calculated?

Hey Nicole, so this confirms my suspicion on the location of the bug. I am going to raise this to the Infra team.

As for the bandwidth, I am going to see if we can waive the period of cost that you are facing, but it doesn't matter the storage in your Redis but the traffic that you are doing to that service, it seems like you are doing a lot of reads over the network hence the cost.

Thanks for the reply!

Hey Nicole, so this confirms my suspicion on the location of the bug. I am going to raise this to the Infra team.

About the IPv6 issue...

We are trying to use Supabase's direct connection today, which is IPv6 only, unfortunately, we failed to connect, the error message indicates the "network unreachable".

I searched the support forum, learned that the Railway is fully IPv6 compatible, was this the same issue you mentioned? Was this the same bug?

As for the bandwidth, I am going to see if we can waive the period of cost that you are facing, but it doesn't matter the storage in your Redis but the traffic that you are doing to that service, it seems like you are doing a lot of reads over the network hence the cost.

Thanks for the comments, we have identified the reading issue from our side, it was indeed caused by the insane amount of full-amount of data reading from Redis.

We have fixed this bug, the bandwidth reduced, and everything looks normal now except the IPv6 internal network.

Hopefully this will not interfere any other services on Railway's side, or customers you have.

And, thanks for the waive check, it would be awesome if we could get some.

For Redis connectivity issue, this is quite interesting, the network was working on our staging environment, but not for the production environment.

There is a known bug where the private network is not attached again after a volume migration, I assume that's what Angelo was getting at.

A redeploy of the database has always fixed this.

So the recommended way of fixing this would be recreating of the Redis instance, right? Is the volume required to be deleted? (we still left gigabytes of data inside it)

Simply redeploy and the problem is fixed, based on my previous message, may I ask how you came to the conclusion that you would need to recreate anything?