-
Notifications
You must be signed in to change notification settings - Fork 283
Server crash due to websocket timeout #2097
Comments
Looks like infrastructure instability is causing crashes more regularly now. Seems like this should just pause any OB activity rather than crash the daemon. Maybe you could also point to other publicly available blockbook servers hosted by other organizations?
|
Thanks for the report @jlopp. We recently migrated to a newer version of BlockBook along with a newer architecture that combines full nodes + BlockBook API on one box, and started having issues with the BTC instances as a result. We're looking into fixing both issues - making the code more resilient to disconnections and the infrastructure more reliable. Will update here. |
I couldn't replicate this issue. I left a couple Dockerised ob-go stores running, one with the same reported version (v0.14.3) pointing at our dev replica of prod that often switches itself off, but ob-go didn't crash at any point. What I did notice however, is that the retry logic seems to be broken - BTC hasn't even retried connecting since the 7th, so this is clearly a separate bug we need to look into (seems it's been raised already as #1520). @jlopp do you have more verbose logs? It would be helpful if you could start the daemon with the Also was this version compiled by yourself, or did you download our binaries? |
@szollo I'm running the amd64 binary; not seeing a verbose logging option. -v just prints the version and exits |
@jlopp sorry, I mean |
Gotcha; running with verbose logging and will update if it crashes again. Might be a few weeks or more... |
Got another crash with verbose logging but not seeing anything particularly interesting...
|
Thanks for the logs @jlopp. As we're not seeing this behaviour on the Dockerised builds of the same version I can only conclude there's an build discrepancy between our Docker images and our binaries. I'm testing this theory now and running the binary the same way you're doing, rather than the Dockerised version of it. In order to fix this now, you might be better off using our Docker images ( In the meantime I'm keeping this open as we figure out what the issue is with the binaries and how to fix it. |
We had a crash of OpenBazaar Server v0.14.3; the final message printed to the console was:
2020/07/22 13:15:37 no response from peer
The tail end of ob.log:
The text was updated successfully, but these errors were encountered: