So Argo Tunnel relies on a single Cloudflare data center?
Argo tunnel generates its DNS zones via the API, and the API is central to one datacenter.
Good to know @Judge . Really hoping this 3+ hr outage due to maintenance was of the emergency kind instead of the scheduled kind during the middle of the day in the middle of the week. I look forward to the postmortem on this one!
This is Cloudflare’s CEO on the issue.
So it was unnecessary scheduled maintenance of a critical service in the middle of the day in the middle of the week then. Fantastic! /s
I don’t believe it was, their original intent, as far as I can tell was to decommission an unused rack of equipment. Their mistake was to not know that there was critical equipment there (a patch panel).
It would be great if the one Cloudflare feature keeping all of my sites offline wasn’t the one that is still down 6 hours later.
What’s especially frustrating as this product that we pay a monthly and per-gigabyte premium for was marketed as the ultimate solution to routing issues, and today we’ve learned that it depends on a single set of API servers living in a single datacenter with a single network uplink and if that fails for any reason the entire system falls apart and every domain using it goes offline.
Absolutely. Loss of API access should not result in loss of existing tunnels.
Nodes rotate regularly, that’s the reason why the loss was gradual. For sure it will be reengineered.
It doesn’t even appear to be hard to do, simply make it so that if the API fails, but the tunnel is up keep it up until you receive a positive reply from the API.
Here is the post-mortem.
And of course they entirely ignore Argo being totally offline and claim that it was just “tunnel creation” that was impacted.
My normal sever ping from Mumbai, India is 128ms.
After using Argo it reduced to around 29-40 ms.
But whenever I on development mode and then off development mode & check the ping it goes to 247 ms from Mumbai India. Site becomes really slow. What to do? Any solution kindly help.
It was, old tunnels are periodically restarted, hence if a tunnel expired during the outage it couldn’t restart a new one.