filters for recent contact-infos when checking for live stake (backport #19204) (#19873)

* filters for recent contact-infos when checking for live stake (#19204)

Contact-infos are saved to disk:
https://github.com/solana-labs/solana/blob/9dfeee299/gossip/src/cluster_info.rs#L1678-L1683

and restored on validator start-up:
https://github.com/solana-labs/solana/blob/9dfeee299/core/src/validator.rs#L450

Staked nodes entries will not expire until an epoch after. So when the
validator checks for online stake it is erroneously picking up
contact-infos restored from disk, which breaks the entire
wait-for-supermajority logic:
https://github.com/solana-labs/solana/blob/9dfeee299/core/src/validator.rs#L1515-L1561

This commit adds an extra check for the age of contact-info entries and
filters out old ones.

(cherry picked from commit 7a789e0763)

# Conflicts:
#	core/src/validator.rs

* removes backport merge conflicts

Co-authored-by: behzad nouri <behzadnouri@gmail.com>

core/src/gossip_service.rs

@@ -245,7 +245,7 @@ fn spy(
             .into_iter()
             .map(|x| x.0)
             .collect::<Vec<_>>();
-        tvu_peers = spy_ref.all_tvu_peers().into_iter().collect::<Vec<_>>();
+        tvu_peers = spy_ref.all_tvu_peers();
 
         let found_node_by_pubkey = if let Some(pubkey) = find_node_by_pubkey {
             all_peers.iter().any(|x| x.id == pubkey)

core/src/validator.rs

@@ -1,6 +1,7 @@
 //! The `validator` module hosts all the validator microservices.
 
-use crate::{
+use {
+    crate::{
         broadcast_stage::BroadcastStageType,
         cache_block_meta_service::{CacheBlockMetaSender, CacheBlockMetaService},
         cluster_info::{
@@ -11,6 +12,7 @@ use crate::{
         completed_data_sets_service::CompletedDataSetsService,
         consensus::{reconcile_blockstore_roots_with_tower, Tower},
         contact_info::ContactInfo,
+        crds_gossip_pull::CRDS_GOSSIP_PULL_CRDS_TIMEOUT_MS,
         gossip_service::GossipService,
         max_slots::MaxSlots,
         optimistically_confirmed_bank_tracker::{
@@ -31,10 +33,10 @@ use crate::{
         tpu::{Tpu, DEFAULT_TPU_COALESCE_MS},
         transaction_status_service::TransactionStatusService,
         tvu::{Sockets, Tvu, TvuConfig},
-};
-use crossbeam_channel::{bounded, unbounded};
-use rand::{thread_rng, Rng};
-use solana_ledger::{
+    },
+    crossbeam_channel::{bounded, unbounded},
+    rand::{thread_rng, Rng},
+    solana_ledger::{
         bank_forks_utils,
         blockstore::{Blockstore, BlockstoreSignals, CompletedSlotsReceiver, PurgeType},
         blockstore_db::BlockstoreRecoveryMode,
@@ -42,17 +44,17 @@ use solana_ledger::{
         leader_schedule::FixedSchedule,
         leader_schedule_cache::LeaderScheduleCache,
         poh::compute_hash_time_ns,
-};
-use solana_measure::measure::Measure;
-use solana_metrics::datapoint_info;
-use solana_runtime::{
+    },
+    solana_measure::measure::Measure,
+    solana_metrics::datapoint_info,
+    solana_runtime::{
         accounts_index::AccountSecondaryIndexes,
         bank::Bank,
         bank_forks::{BankForks, SnapshotConfig},
         commitment::BlockCommitmentCache,
         hardened_unpack::{open_genesis_config, MAX_GENESIS_ARCHIVE_UNPACKED_SIZE},
-};
-use solana_sdk::{
+    },
+    solana_sdk::{
         clock::Slot,
         epoch_schedule::MAX_LEADER_SCHEDULE_EPOCH_OFFSET,
         genesis_config::GenesisConfig,
@@ -61,20 +63,23 @@ use solana_sdk::{
         shred_version::compute_shred_version,
         signature::{Keypair, Signer},
         timing::timestamp,
-};
-use solana_vote_program::vote_state::VoteState;
-use std::time::Instant;
-use std::{
-    collections::HashSet,
+    },
+    solana_vote_program::vote_state::VoteState,
+    std::time::Instant,
+    std::{
+        collections::{HashMap, HashSet},
         fmt,
         net::SocketAddr,
         ops::Deref,
         path::{Path, PathBuf},
-    sync::atomic::{AtomicBool, AtomicU64, Ordering},
-    sync::mpsc::Receiver,
-    sync::{Arc, Mutex, RwLock},
+        sync::{
+            atomic::{AtomicBool, AtomicU64, Ordering},
+            mpsc::Receiver,
+            Arc, Mutex, RwLock,
+        },
         thread::{sleep, Builder},
         time::Duration,
+    },
 };
 
 const MAX_COMPLETED_DATA_SETS_IN_CHANNEL: usize = 100_000;
@@ -1458,7 +1463,20 @@ fn get_stake_percent_in_gossip(bank: &Bank, cluster_info: &ClusterInfo, log: boo
     let mut offline_nodes = vec![];
 
     let mut total_activated_stake = 0;
-    let all_tvu_peers = cluster_info.all_tvu_peers();
+    let now = timestamp();
+    // Nodes contact infos are saved to disk and restored on validator startup.
+    // Staked nodes entries will not expire until an epoch after. So it
+    // is necessary here to filter for recent entries to establish liveness.
+    let peers: HashMap<_, _> = cluster_info
+        .all_tvu_peers()
+        .into_iter()
+        .filter(|node| {
+            let age = now.saturating_sub(node.wallclock);
+            // Contact infos are refreshed twice during this period.
+            age < CRDS_GOSSIP_PULL_CRDS_TIMEOUT_MS
+        })
+        .map(|node| (node.id, node))
+        .collect();
     let my_shred_version = cluster_info.my_shred_version();
     let my_id = cluster_info.id();
 
@@ -1474,10 +1492,7 @@ fn get_stake_percent_in_gossip(bank: &Bank, cluster_info: &ClusterInfo, log: boo
             .map(|vote_state| vote_state.node_pubkey)
             .unwrap_or_default();
 
-        if let Some(peer) = all_tvu_peers
-            .iter()
-            .find(|peer| peer.id == vote_state_node_pubkey)
-        {
+        if let Some(peer) = peers.get(&vote_state_node_pubkey) {
             if peer.shred_version == my_shred_version {
                 trace!(
                     "observed {} in gossip, (activated_stake={})",