Describe the bug
While reviewing #781 I found a bug that could exhaust device resources for long runs, I decided to write this issue instead of a pr comment because it got too verbose.
Seeing that push_if_has_service is being called 3 times in a row for the same LocalAddress i started thinking if there was any dedup going on... I saw that push_addresses does check duplicates but push_if_has_service does not.
so i added the below here on address_man:422 inside the prune_addresses method, which is called evey push_address
// Print duplicates
for (service, peers) in &self.peers_by_service { // iterate over piers
// cast them into a HashSet, a hashset should not retain any duplicates.
let unique: std::collections::HashSet<&usize> = peers.iter().collect();
// print when their length mismatch
if peers.len() != unique.len() {
println!(
" DUPLICATES in peers_by_service[{:?}]: total={} unique={}",
service,
peers.len(),
unique.len()
);
}
}
for (service, peers) in &self.good_peers_by_service {
// cast them into a HashSet, a hashset should not retain any duplicates.
let unique: std::collections::HashSet<&usize> = peers.iter().collect();
// print when their length mismatch
if peers.len() != unique.len() {
println!(
" DUPLICATES in good_peers_by_service[{:?}]: total={} unique={}",
service,
peers.len(),
unique.len()
);
}
}
println!(
"prune_addresses called: total={} max={} excess={}",
self.addresses.len(),
self.max_size,
excess
);When running florestad in a normal run, e.g. cargo run --bin florestad, i could already see:
2026-02-21 17:57:17 INFO floresta_wire::p2p_wire::node::chain_selector_ctx: Downloading headers from peer=0 at height=24001 hash=00000000057136237ef09824c432d8151cc23bcb009cecf49ace17003fe574be
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=1742 unique=1740
prune_addresses called: total=1740 max=50000 excess=0
2026-02-21 17:57:17 ERROR floresta_wire::p2p_wire::node::chain_selector_ctx: 798: crates/floresta-wire/src/p2p_wire/node/chain_selector_ctx.rs - NoPeersAvailable
2026-02-21 17:57:38 INFO floresta_wire::p2p_wire::node::peer_man: New peer id=4 version=/Satoshi:30.0.0/ blocks=937711 services=ServiceFlags(NETWORK|WITNESS|COMPACT_FILTERS|NETWORK_LIMITED|P2P_V2)
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=1064 unique=1063
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=2300 unique=2296
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=358 unique=357
prune_addresses called: total=2296 max=50000 excess=0
2026-02-21 17:58:00 INFO floresta_wire::p2p_wire::node::peer_man: New peer id=7 version=/Satoshi:30.0.0/ blocks=937711 services=ServiceFlags(NETWORK|BLOOM|WITNESS|COMPACT_FILTERS|NETWORK_LIMITED|P2P_V2)
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=1301 unique=1299
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=2828 unique=2823
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=357 unique=356
prune_addresses called: total=2823 max=50000 excess=0
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=1302 unique=1300
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=2830 unique=2825
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=357 unique=356
prune_addresses called: total=2825 max=50000 excess=0
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=1302 unique=1300
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=2830 unique=2825
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=356 unique=355
prune_addresses called: total=2825 max=50000 excess=0
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=1303 unique=1301
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=2831 unique=2826
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=355 unique=354
prune_addresses called: total=2826 max=50000 excess=0
the duplicates continue to grow:
prune_addresses called: total=8252 max=50000 excess=0
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=3673 unique=3665
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=8268 unique=8252
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=1119 unique=1117
DUPLICATES in good_peers_by_service[ServiceFlags(64)]: total=453 unique=452
prune_addresses called: total=8252 max=50000 excess=0
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=3673 unique=3665
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=8268 unique=8252
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=1119 unique=1117
DUPLICATES in good_peers_by_service[ServiceFlags(64)]: total=453 unique=452
prune_addresses called: total=8252 max=50000 excess=0
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=3673 unique=3665
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=8268 unique=8252
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=1119 unique=1117
DUPLICATES in good_peers_by_service[ServiceFlags(64)]: total=453 unique=452
prune_addresses called: total=8252 max=50000 excess=0
DUPLICATES in peers_by_service[ServiceFlags(64)]: total=3673 unique=3665
DUPLICATES in peers_by_service[ServiceFlags(0)]: total=8268 unique=8252
DUPLICATES in good_peers_by_service[ServiceFlags(0)]: total=1119 unique=1117
DUPLICATES in good_peers_by_service[ServiceFlags(64)]: total=453 unique=452
prune_addresses called: total=8252 max=50000 excess=0
therefore, when i add the code below on address_man:511 at the beginning of push_if_has_service.
// don't add duplicate addresses
if self
.addresses
.values()
.any(|x| x.address == address.address)
{
return;
}And by restarting florestad I could see the behavior to change:
2026-02-21 18:21:08 INFO floresta_wire::p2p_wire::node::conn: Fetched 347 peer addresses from all DNS seeds
prune_addresses called: total=8299 max=50000 excess=0
2026-02-21 18:21:07 INFO floresta_wire::p2p_wire::address_man: Fetched 39 peer addresses from DNS host: x9.dnsseed.emzy.de
2026-02-21 18:21:08 INFO floresta_wire::p2p_wire::address_man: Fetched 31 peer addresses from DNS host: x49.seed.bitcoin.wiz.biz
2026-02-21 18:21:08 INFO floresta_wire::p2p_wire::address_man: Fetched 31 peer addresses from DNS host: x9.seed.bitcoin.wiz.biz
2026-02-21 18:21:08 INFO floresta_wire::p2p_wire::node::conn: Fetched 347 peer addresses from all DNS seeds
prune_addresses called: total=8299 max=50000 excess=0
22026-02-21 18:22:24 INFO floresta_wire::p2p_wire::node::peer_man: New peer id=9 version=/Satoshi:30.2.0/ blocks=937712 services=ServiceFlags(NETWORK|WITNESS|NETWORK_LIMITED|P2P_V2)
prune_addresses called: total=8984 max=50000 excess=0
2026-02-21 18:22:25 INFO floresta_wire::p2p_wire::node::chain_selector_ctx: Downloading headers from peer=9 at height=26001 hash=000000006aee625129fc9ad7ba7930b89c80ebf22ce4448d5ebdd8609fdf0a3d
Describe the bug
While reviewing #781 I found a bug that could exhaust device resources for long runs, I decided to write this issue instead of a pr comment because it got too verbose.
Seeing that
push_if_has_serviceis being called 3 times in a row for the sameLocalAddressi started thinking if there was any dedup going on... I saw thatpush_addressesdoes check duplicates butpush_if_has_servicedoes not.so i added the below here on
address_man:422inside theprune_addressesmethod, which is called eveypush_addressWhen running florestad in a normal run, e.g.
cargo run --bin florestad, i could already see:the duplicates continue to grow:
therefore, when i add the code below on
address_man:511at the beginning ofpush_if_has_service.And by restarting
florestadI could see the behavior to change: