Beacon client online, but no consensus updates received

[ncavedale-xlabs]

Hi team!
Every now and then, the node stops syncing with the following message in geth logs:

Oct 30 20:36:14 geth[248526]: WARN [10-30|20:36:14.391] Beacon client online, but no consensus updates received in a while. Please fix your beacon client to follow the chain!

And I don't see anything unusual in the beacon logs. The solution is to simply restart the beacond service. It starts working fine after that.

Were you aware of this?
Perhaps there's a beacon config setting that can help in these cases?

[chuck-bear]

We might need more information here. is that a fresh install? Is that a node using a snapshot? What's the beacond version?

But first, can you compare your beacon config with the toml files listed here: https://github.com/berachain/beacon-kit/tree/main/testing/networks/80084

[ncavedale-xlabs]

This is not a fresh install, but it has also happened before when we were running reth (we're running geth now).
We used a snapshot for beacon but not for geth. The beacon version is "v0.2.0-alpha.8".
I compared the .toml files: the config.toml is the same we're already using, but there are some differences in app.toml:

app.toml in repo:

rpc-timeout = "900ms"
rpc-jwt-refresh-interval = "20s"

app.toml in our server:

rpc-timeout = "2s"
rpc-jwt-refresh-interval = "30s"

Also noticed that sections like "API Configuration" and "gRPC Configuration" are no longer present, so I can try using a new copy of app.toml to see if that helps.

FYI, we're also running with a patched binary: #2029

[chuck-bear]

This is a known issue:
rpc-timeout = "2s"

[chuck-bear]

And you issue was linked to the rpc-timeout, no need to run a patched version.

[ncavedale-xlabs]

The default timeout (on code) was 2s, and by patching the binary we increased it to 3s. If I set it to 900ms like in app.toml, wouldn't that cause the issue to reappear?

In any case, that was a different issue where beacon would restart over and over again. Our current issue however, seems to be a stop in the communication between beacon and geth/reth. What would you suggest for this?

[ncavedale-xlabs]

We're now using a fresh copy of the app.toml file and have re-compiled the beacon binary to get rid of the patch for the previous issue.
For now the node is running fine 👍