Ivan Kozlovic

It looks to me that the server - as soon as routes are created - is overwhelmed with traffic to the point that the internal connections requests to the NATS...

> So how can I deal with it? I described some of the ways we could try to debug this by capturing some of the traffic and see if this...

> I think the problem occurs in cases where multiple clients are trying to connect to the server (establish new connection) at the same time. If this is following a...

If you start with streaming "store" but no RAFT logs, those may be created empty but the system will detect that there was no RAFT state but a streaming store....

If you have the remaining of the cluster working fine, say 2 out of 3 nodes running with a new leader, technically you could remove the state from the failed...

That means the raft logs are corrupted, so in that case you have no choice but to start that node without any store (raft+streaming). It should be able to ultimately...

That's tricky. So definitively you should save off the streaming stores (where the messages are), just in case the restarted node cannot fully recover before one of the 2 other...

I would be a bit scared of automatically removing those files. That would not give a chance to the user to make a copy.

The fact that it says: ``` [1] 2020/04/19 22:49:13.407013 [INF] STREAM: Recovered 0 channel(s) ``` indicates to me that the directory already existed. Otherwise you would have had something like:...

Puzzling indeed, but the error you describe would mean really that the streaming store was found while the raft was not. Of course once streaming and raft stores are created...