What happens when half of the IPFS network is down?

blog.ipfs.tech

87 points·dennis-tra··9 comments

>The problematic configuration which was applied manually (i.e. was not based on the default values of kubo-v0.17)

So how did 60% of the network become manually mis-configured? Or does a single company (cloudflare?) make up 60% of the network?

From my reading, they mis-configured it in their use of the library - using the default value instead would have avoided the problem.

Someone did some digging on various non-Bitcoin crypto coins and found that a huge percentage of the whatever it is they run were just running on AWS.

I guess you mean nodes?

I only found this, which is about eth: https://www.reddit.com/r/Buttcoin/comments/z4e4g1/60_eth_nod...

Worth correcting: 59% of hosted nodes use Amazon, or 37% of all nodes. If you filter out unsynced nodes, about 17% of all nodes are hosted on Amazon.

https://ethernodes.org/networkType/Hosting?synced=1

That looks like it might be what I saw; of course all the tiny cryptos are run on only a few machines, but the cloud providers are huge.

Like the entire application binary was mis-configured. so anyone running the newest update would be misconfigured

Am I the only one who finds it a bit odd that the article starts off bragging about what was an outage?

And I as soon as I hear “we were still running with half our network down”, the first thing I think of is that no one really is using IPFS at all.

"Connection was reset."

Well, that might address the mystery :-)

Reply on news.ycombinator.com