peer table thresholds

This commit is contained in:
John Smith 2022-03-24 10:14:50 -04:00
parent ac0280e0b6
commit c276dd7796
10 changed files with 184 additions and 47 deletions

View File

@ -188,7 +188,7 @@ struct SenderInfo {
struct OperationInfoA {
nodeInfo @0 :NodeInfo; # returned node information
senderInfo @1 :SenderInfo; # info about InfoQ sender
senderInfo @1 :SenderInfo; # info about InfoQ sender from the perspective of the replier
}
struct OperationValidateDialInfo {
@ -204,7 +204,7 @@ struct OperationReturnReceipt {
struct OperationFindNodeQ {
nodeId @0 :NodeID; # node id to locate
peerInfo @1 :PeerInfo; # The peer info for node asking the question
dialInfoList @1 :List(DialInfo); # dial info for the node asking the question
}
struct PeerInfo {
@ -368,7 +368,7 @@ struct Operation {
respondTo :union {
none @1 :Void; # no response is desired
sender @2 :Void; # envelope sender node id to be used for reply
sender @2 :DialInfo; # (Optional) envelope sender node id to be used for reply
# possibly through a relay if the request arrived that way
privateRoute @3 :PrivateRoute; # embedded private route to be used for reply
}

View File

@ -2,6 +2,7 @@ use crate::callback_state_machine::*;
use crate::dht::crypto::Crypto;
use crate::intf::*;
use crate::network_manager::*;
use crate::routing_table::*;
use crate::xx::*;
use crate::*;
use core::convert::TryFrom;
@ -106,8 +107,8 @@ pub struct AttachmentManagerInner {
attachment_machine: CallbackStateMachine<Attachment>,
network_manager: NetworkManager,
maintain_peers: bool,
peer_count: u32,
attach_timestamp: Option<u64>,
update_callback: Option<UpdateCallback>,
attachment_maintainer_jh: Option<JoinHandle<()>>,
}
@ -127,8 +128,8 @@ impl AttachmentManager {
attachment_machine: CallbackStateMachine::new(),
network_manager: NetworkManager::new(config, table_store, crypto),
maintain_peers: false,
peer_count: 0,
attach_timestamp: None,
update_callback: None,
attachment_maintainer_jh: None,
}
}
@ -159,24 +160,30 @@ impl AttachmentManager {
self.inner.lock().attach_timestamp
}
pub fn get_peer_count(&self) -> u32 {
self.inner.lock().peer_count
}
fn translate_peer_input(cur: u32, max: u32) -> AttachmentInput {
if cur > max {
fn translate_routing_table_health(
health: RoutingTableHealth,
config: &VeilidConfigRoutingTable,
) -> AttachmentInput {
if health.reliable_entry_count >= config.limit_over_attached.try_into().unwrap() {
return AttachmentInput::TooManyPeers;
}
match cmp::min(4, 4 * cur / max) {
4 => AttachmentInput::FullPeers,
3 => AttachmentInput::StrongPeers,
2 => AttachmentInput::GoodPeers,
1 => AttachmentInput::WeakPeers,
0 => AttachmentInput::NoPeers,
_ => panic!("Invalid state"),
if health.reliable_entry_count >= config.limit_fully_attached.try_into().unwrap() {
return AttachmentInput::FullPeers;
}
if health.reliable_entry_count >= config.limit_attached_strong.try_into().unwrap() {
return AttachmentInput::StrongPeers;
}
fn translate_peer_state(state: &AttachmentState) -> AttachmentInput {
if health.reliable_entry_count >= config.limit_attached_good.try_into().unwrap() {
return AttachmentInput::GoodPeers;
}
if health.reliable_entry_count >= config.limit_attached_weak.try_into().unwrap()
|| health.unreliable_entry_count >= config.limit_attached_weak.try_into().unwrap()
{
return AttachmentInput::WeakPeers;
}
AttachmentInput::NoPeers
}
fn translate_attachment_state(state: &AttachmentState) -> AttachmentInput {
match state {
AttachmentState::OverAttached => AttachmentInput::TooManyPeers,
AttachmentState::FullyAttached => AttachmentInput::FullPeers,
@ -188,19 +195,20 @@ impl AttachmentManager {
}
}
async fn update_peer_count(&self) {
async fn update_attachment(&self) {
let new_peer_state_input = {
let inner = self.inner.lock();
let old_peer_state_input =
AttachmentManager::translate_peer_state(&inner.attachment_machine.state());
AttachmentManager::translate_attachment_state(&inner.attachment_machine.state());
let max_connections = inner.config.get().network.max_connections;
// get active peer count from routing table
// get reliable peer count from routing table
let routing_table = inner.network_manager.routing_table();
let health = routing_table.get_routing_table_health();
let routing_table_config = &inner.config.get().network.routing_table;
let new_peer_state_input =
AttachmentManager::translate_peer_input(inner.peer_count, max_connections);
AttachmentManager::translate_routing_table_health(health, routing_table_config);
if old_peer_state_input == new_peer_state_input {
None
@ -238,8 +246,7 @@ impl AttachmentManager {
break;
}
// xxx: ?update peer count?
self.update_peer_count().await;
self.update_attachment().await;
// sleep should be at the end in case maintain_peers changes state
intf::sleep(1000).await;
@ -259,16 +266,16 @@ impl AttachmentManager {
self.inner.lock().attach_timestamp = None;
}
pub async fn init(
&self,
state_change_callback: StateChangeCallback<Attachment>,
) -> Result<(), String> {
pub async fn init(&self, update_callback: UpdateCallback) -> Result<(), String> {
trace!("init");
let network_manager = {
let inner = self.inner.lock();
inner
.attachment_machine
.set_state_change_callback(state_change_callback);
let mut inner = self.inner.lock();
inner.update_callback = Some(update_callback.clone());
inner.attachment_machine.set_state_change_callback(Arc::new(
move |_old_state: AttachmentState, new_state: AttachmentState| {
update_callback(VeilidUpdate::Attachment { state: new_state })
},
));
inner.network_manager.clone()
};
@ -284,6 +291,8 @@ impl AttachmentManager {
inner.network_manager.clone()
};
network_manager.terminate().await;
let mut inner = self.inner.lock();
inner.update_callback = None;
}
fn attach(&self) {

View File

@ -70,7 +70,7 @@ impl ConnectionManager {
pub async fn startup(&self) {
trace!("startup connection manager");
let mut inner = self.arc.inner.lock().await;
let cac = flume::bounded(CONNECTION_PROCESSOR_CHANNEL_SIZE); // xxx move to config
let cac = flume::bounded(CONNECTION_PROCESSOR_CHANNEL_SIZE);
inner.connection_add_channel_tx = Some(cac.0);
let rx = cac.1.clone();
let this = self.clone();

View File

@ -115,16 +115,9 @@ impl ServicesContext {
// Set up attachment manager
trace!("init attachment manager");
let update_callback_move = self.update_callback.clone();
let update_callback = self.update_callback.clone();
let attachment_manager = AttachmentManager::new(self.config.clone(), table_store, crypto);
if let Err(e) = attachment_manager
.init(Arc::new(
move |_old_state: AttachmentState, new_state: AttachmentState| {
update_callback_move(VeilidUpdate::Attachment { state: new_state })
},
))
.await
{
if let Err(e) = attachment_manager.init(update_callback).await {
self.shutdown().await;
return Err(VeilidAPIError::Internal { message: e });
}

View File

@ -60,6 +60,13 @@ struct RoutingTableInner {
self_transfer_stats: TransferStatsDownUp,
}
#[derive(Clone, Debug, Default)]
pub struct RoutingTableHealth {
pub reliable_entry_count: usize,
pub unreliable_entry_count: usize,
pub dead_entry_count: usize,
}
struct RoutingTableUnlockedInner {
// Background processes
rolling_transfers_task: TickTask,
@ -743,4 +750,29 @@ impl RoutingTable {
e.question_lost(ts);
})
}
//////////////////////////////////////////////////////////////////////
// Routing Table Health Metrics
pub fn get_routing_table_health(&self) -> RoutingTableHealth {
let mut health = RoutingTableHealth::default();
let cur_ts = intf::get_timestamp();
let inner = self.inner.lock();
for bucket in &inner.buckets {
for entry in bucket.entries() {
match entry.1.state(cur_ts) {
BucketEntryState::Reliable => {
health.reliable_entry_count += 1;
}
BucketEntryState::Unreliable => {
health.unreliable_entry_count += 1;
}
BucketEntryState::Dead => {
health.dead_entry_count += 1;
}
}
}
}
health
}
}

View File

@ -191,6 +191,11 @@ fn config_callback(key: String) -> ConfigCallbackReturn {
"network.node_id" => Ok(Box::new(dht::key::DHTKey::default())),
"network.node_id_secret" => Ok(Box::new(dht::key::DHTKeySecret::default())),
"network.bootstrap" => Ok(Box::new(Vec::<String>::new())),
"network.routing_table.limit_over_attached" => Ok(Box::new(64u32)),
"network.routing_table.limit_fully_attached" => Ok(Box::new(32u32)),
"network.routing_table.limit_attached_strong" => Ok(Box::new(16u32)),
"network.routing_table.limit_attached_good" => Ok(Box::new(8u32)),
"network.routing_table.limit_attached_weak" => Ok(Box::new(4u32)),
"network.rpc.concurrency" => Ok(Box::new(2u32)),
"network.rpc.queue_size" => Ok(Box::new(128u32)),
"network.rpc.max_timestamp_behind_ms" => Ok(Box::new(Some(10_000u32))),
@ -300,6 +305,12 @@ pub async fn test_config() {
assert_eq!(inner.network.rpc.queue_size, 128u32);
assert_eq!(inner.network.rpc.timeout_ms, 10_000u32);
assert_eq!(inner.network.rpc.max_route_hop_count, 7u8);
assert_eq!(inner.network.routing_table.limit_over_attached, 64u32);
assert_eq!(inner.network.routing_table.limit_fully_attached, 32u32);
assert_eq!(inner.network.routing_table.limit_attached_strong, 16u32);
assert_eq!(inner.network.routing_table.limit_attached_good, 8u32);
assert_eq!(inner.network.routing_table.limit_attached_weak, 4u32);
assert_eq!(
inner.network.dht.resolve_node_timeout_ms,
Option::<u32>::None

View File

@ -104,6 +104,7 @@ pub struct VeilidConfigDHT {
pub min_peer_count: u32,
pub min_peer_refresh_time_ms: u32,
pub validate_dial_info_receipt_time_ms: u32,
pub nearby_node_percentage: u32,
}
#[derive(Default, Clone, Serialize, Deserialize)]
@ -123,6 +124,14 @@ pub struct VeilidConfigLeases {
pub max_client_signal_leases: u32,
pub max_client_relay_leases: u32,
}
#[derive(Default, Clone, Serialize, Deserialize)]
pub struct VeilidConfigRoutingTable {
pub limit_over_attached: u32,
pub limit_fully_attached: u32,
pub limit_attached_strong: u32,
pub limit_attached_good: u32,
pub limit_attached_weak: u32,
}
#[derive(Default, Clone, Serialize, Deserialize)]
pub struct VeilidConfigNetwork {
@ -132,6 +141,7 @@ pub struct VeilidConfigNetwork {
pub node_id: key::DHTKey,
pub node_id_secret: key::DHTKeySecret,
pub bootstrap: Vec<String>,
pub routing_table: VeilidConfigRoutingTable,
pub rpc: VeilidConfigRPC,
pub dht: VeilidConfigDHT,
pub upnp: bool,
@ -285,6 +295,11 @@ impl VeilidConfig {
get_config!(inner.network.connection_initial_timeout_ms);
get_config!(inner.network.connection_inactivity_timeout_ms);
get_config!(inner.network.bootstrap);
get_config!(inner.network.routing_table.limit_over_attached);
get_config!(inner.network.routing_table.limit_fully_attached);
get_config!(inner.network.routing_table.limit_attached_strong);
get_config!(inner.network.routing_table.limit_attached_good);
get_config!(inner.network.routing_table.limit_attached_weak);
get_config!(inner.network.dht.resolve_node_timeout_ms);
get_config!(inner.network.dht.resolve_node_count);
get_config!(inner.network.dht.resolve_node_fanout);

View File

@ -44,6 +44,13 @@ Future<VeilidConfig> getDefaultVeilidConfig() async {
nodeId: "",
nodeIdSecret: "",
bootstrap: [],
routingTable: VeilidConfigRoutingTable(
limitOverAttached: 64,
limitFullyAttached: 32,
limitAttachedStrong: 16,
limitAttachedGood: 8,
limitAttachedWeak: 4,
),
rpc: VeilidConfigRPC(
concurrency: 0,
queueSize: 1024,

View File

@ -483,6 +483,41 @@ class VeilidConfigRPC {
////////////
class VeilidConfigRoutingTable {
int limitOverAttached;
int limitFullyAttached;
int limitAttachedStrong;
int limitAttachedGood;
int limitAttachedWeak;
VeilidConfigRoutingTable({
required this.limitOverAttached,
required this.limitFullyAttached,
required this.limitAttachedStrong,
required this.limitAttachedGood,
required this.limitAttachedWeak,
});
Map<String, dynamic> get json {
return {
'limit_over_attached': limitOverAttached,
'limit_fully_attached': limitFullyAttached,
'limit_attached_strong': limitAttachedStrong,
'limit_attached_good': limitAttachedGood,
'limit_attached_weak': limitAttachedWeak,
};
}
VeilidConfigRoutingTable.fromJson(Map<String, dynamic> json)
: limitOverAttached = json['limit_over_attached'],
limitFullyAttached = json['limit_fully_attached'],
limitAttachedStrong = json['limit_attached_strong'],
limitAttachedGood = json['limit_attached_good'],
limitAttachedWeak = json['limit_attached_weak'];
}
////////////
class VeilidConfigLeases {
int maxServerSignalLeases;
int maxServerRelayLeases;
@ -520,6 +555,7 @@ class VeilidConfigNetwork {
String nodeId;
String nodeIdSecret;
List<String> bootstrap;
VeilidConfigRoutingTable routingTable;
VeilidConfigRPC rpc;
VeilidConfigDHT dht;
bool upnp;
@ -538,6 +574,7 @@ class VeilidConfigNetwork {
required this.nodeId,
required this.nodeIdSecret,
required this.bootstrap,
required this.routingTable,
required this.rpc,
required this.dht,
required this.upnp,
@ -558,6 +595,7 @@ class VeilidConfigNetwork {
'node_id': nodeId,
'node_id_secret': nodeIdSecret,
'bootstrap': bootstrap,
'routing_table': routingTable.json,
'rpc': rpc.json,
'dht': dht.json,
'upnp': upnp,
@ -579,6 +617,7 @@ class VeilidConfigNetwork {
nodeId = json['node_id'],
nodeIdSecret = json['node_id_secret'],
bootstrap = json['bootstrap'],
routingTable = VeilidConfigRoutingTable.fromJson(json['routing_table']),
rpc = VeilidConfigRPC.fromJson(json['rpc']),
dht = VeilidConfigDHT.fromJson(json['dht']),
upnp = json['upnp'],

View File

@ -54,6 +54,12 @@ core:
node_id: ''
node_id_secret: ''
bootstrap: []
routing_table:
limit_over_attached: 64
limit_fully_attached: 32
limit_attached_strong: 16
limit_attached_good: 8
limit_attached_weak: 4
rpc:
concurrency: 0
queue_size: 1024
@ -520,6 +526,15 @@ pub struct Leases {
pub max_client_relay_leases: u32,
}
#[derive(Debug, Deserialize, Serialize)]
pub struct RoutingTable {
pub limit_over_attached: u32,
pub limit_fully_attached: u32,
pub limit_attached_strong: u32,
pub limit_attached_good: u32,
pub limit_attached_weak: u32,
}
#[derive(Debug, Deserialize, Serialize)]
pub struct Network {
pub max_connections: u32,
@ -528,6 +543,7 @@ pub struct Network {
pub node_id: veilid_core::DHTKey,
pub node_id_secret: veilid_core::DHTKeySecret,
pub bootstrap: Vec<ParsedNodeDialInfo>,
pub routing_table: RoutingTable,
pub rpc: Rpc,
pub dht: Dht,
pub upnp: bool,
@ -811,6 +827,21 @@ impl Settings {
.map(|e| e.node_dial_info_string)
.collect::<Vec<String>>(),
)),
"network.routing_table.limit_over_attached" => Ok(Box::new(
inner.core.network.routing_table.limit_over_attached,
)),
"network.routing_table.limit_fully_attached" => Ok(Box::new(
inner.core.network.routing_table.limit_fully_attached,
)),
"network.routing_table.limit_attached_strong" => Ok(Box::new(
inner.core.network.routing_table.limit_attached_strong,
)),
"network.routing_table.limit_attached_good" => Ok(Box::new(
inner.core.network.routing_table.limit_attached_good,
)),
"network.routing_table.limit_attached_weak" => Ok(Box::new(
inner.core.network.routing_table.limit_attached_weak,
)),
"network.rpc.concurrency" => Ok(Box::new(inner.core.network.rpc.concurrency)),
"network.rpc.queue_size" => Ok(Box::new(inner.core.network.rpc.queue_size)),
"network.rpc.max_timestamp_behind_ms" => {