Open maltekliemann opened 1 month ago
The first message is being tracked here: https://github.com/paritytech/polkadot-sdk/issues/1915 and here: https://github.com/paritytech/polkadot-sdk/issues/531
There has been recent work to avoid these duplicate block requests: https://github.com/paritytech/polkadot-sdk/pull/5029 However, that fix is fairly recent and will not come to 1.1.0. The issue should not appear too often however.
Regarding the grandpa neighbor message I am not sure. cc @lexnv
Thanks for the reply. Am I interpreting this correctly: We don't have to worry about halting our network due to at least the first message?
We don't have to worry about halting our network due to at least the first message?
I would say everything is ok, nothing to worry about from those 2 log messages. Generally, we care more about about block production and connectivity to multiple peers. It is ok for some peers to get disconnected, especially if they "misbehave".
We currently ban peers that have made the same request to us multiple time (3 times).
This is causing the first error message Same block request
.
This PR https://github.com/paritytech/polkadot-sdk/pull/5029 aims to offload some pressure from peers who are slow to respond. However, we'd have to wait a bit for it to get deployed to the majority of the network before we see some significant effects.
The second message is not concerning. We are adding +100
to the reputation of a banned peer. This is expected behavior and we are "rewarding" the peer by increasing its reputation.
We are emitting warnings if the reputation of a peer gets under a threshold. In this case, the +100 did not increase the peer's reputation enough to escape the banned threshold.
We have two bootnodes running at polkadot-v0.9.38. Upgrading one of these to our new client, which is at polkadot-v1.1.0 raises the following two error messages:
(The runtime is still at the old version at this point.)
We have failed to reproduce these error messages using a non-bootnode client on our local machines, also in integration tests or using a local parachain network - no errors!
What causes these errors? Are they bootnode related? Can we expect them to vanish once a majority of the nodes are updated to the new client? Should we be concerned about going through with this update?