Crash during catchup with trusted hash about casper-node

[XeZZoR]

Ah I got it sorry I misread your comment above. That is very possible and would be inline with the log message too.

[XeZZoR]

Attaching source Setup. To reproduce:

Change the trusted hash in config/config.toml

docker volume create clkeys
./rebuild.sh
docker logs -f capser

This should build the container and show logs. The image build takes quite a while and if the era switches in between the trusted hash is invalid to catchup. In that case just exchange the hash and run ./rebuild.sh again it nukes the state completely despite the generated keys.
casperdocker.zip

[afck]

My theory is that some node in the network was reusing a storage directory from a previous network. That's the only way I could reproduce something akin to this issue. #508 (merged into the master branch) should address this: If a peer sends us a block with the wrong parent, we retry with another peer.
(That also means that if we are the ones joining the network with outdated storage, the node would give up after a while: Consensus shutting down due to inability to participate in the network.)

[XeZZoR]

I am very confident this is not the case since I ran it on a complete new server and was able to reproduce it.

[afck]

Yes, but another node in the network could have used an oudated storage directory, and sent an invalid block to your node?