feat(share/p2p/peer-manager): use shrexSub peers as full nodes #2105
Conversation
walldiss
requested review from
renaynay,
Wondertan,
vgonkivs and
distractedm1nd
as code owners
walldiss
force-pushed
the
use_shrex_peer_for_full_nodes
branch
2 times, most recently
from
1e3c701
to
b0ce602
Compare
|
Let's add more description as to why we are doing this. It's pretty important protocol change and while we don't have an issue, we should document the change on the PR |
|
All tests failing here @walldiss |
|
I second what @Wondertan said, lets convert to draft pls @walldiss 🙏🏻 |
|
@Wondertan @renaynay Are you worried about any potential negative effects that may arise from the changes that have been implemented? |
|
@walldiss, I am worried about the added complexity of something that is optional and is not part of the original design. We should add if it's proven to be beneficial. |
distractedm1nd
force-pushed
the
use_shrex_peer_for_full_nodes
branch
from
dfb6355
to
f1f2af1
Compare
|
Let's rebase this |
|
A thought. I am still worried that we have two places where we keep the current state of FN we rely on. This PR makes this discrepancy even worse. Discovery will have its own peer set and PeerManager, but this time they are not just copied but different. The constancies seem to me obvious, but I can explain if needed. I think that during future refactors, we should design shrex in a monolithic way. Where it has multiple coupled and internalized(via |
distractedm1nd
force-pushed
the
use_shrex_peer_for_full_nodes
branch
from
f1f2af1
to
5d740c9
Compare
distractedm1nd
force-pushed
the
use_shrex_peer_for_full_nodes
branch
from
eccffb8
to
c952803
Compare
…g peer manager cooldown issue
…g peer manager cooldown issue
|
✅ LGTM! Can't approve |
Fixes celestiaorg#2204 Regarding the case where we exhausted all the peers on the network and backed them all off. It can happen on the smaller network, and in the worst case, the node will wait 10 minutes(the default backoff time), which is acceptable considering peers from shrexsub celestiaorg#2105
…tiaorg#2105) ## Problem celestiaorg#2107 ## Overview Could hotfix the problem by allowing peer manager to save nodes discovered through shrexSub to be used in full nodes pool. The solution would also benefit shrex performance by allowing distribution of request to more peers. --------- Co-authored-by: Ryan <ryan@celestia.org> Co-authored-by: Hlib Kanunnikov <hlibwondertan@gmail.com>
Problem
#2107
Overview
Could hotfix the problem by allowing peer manager to save nodes discovered through shrexSub to be used in full nodes pool.
The solution would also benefit shrex performance by allowing distribution of request to more peers.