Bro I test GreenCloudVPS last month singapore location, v6 route to my ISP in istanbul go through hong kong first then tokyo then back, latency 380ms. v4 is 180ms direct. This is "working" technically but practically broken bro.
Their support say "v6 path is not guaranteed" lol. So even when provider "support" v6, routing quality is gamble.
I think this is part of regression too, not just on/off switch.
Member
This is exactly the maintenance culture problem I mentioned. GreenCloudVPS gives you the address but their upstream or peering for v6 is afterthought.
I did traceroute comparison last year for client project, wrote it up. Same pattern: v6 often takes scenic route because nobody optimises it. My test to Leaseweb goes via Cogent v6 but Tata v4, 40ms difference.
Here is practical list from my notes:
For production I now dual-stack with v6 as failover only. Sad but realistic.
Same. My monitoring shows v6 failure rate 4x higher than v4 across the 52 boxes. Not total outages, but enough that alerting would cry wolf.
I am considering dropping v6 from production monitoring entirely and just checking monthly. Which defeats the purpose.
Question for thread: anyone using v6-only internally with NAT64 for egress? Wondering if that is saner path than hoping providers fix their dual-stack.
