Add missed blocks to monitored validators (#4731)

* add missed_block metric

* init missed_block in constructor

* declare beaconproposercache in ValidatorMonitor

* refacto proposer_shuffling_decision_root to use epoch instead of current.epoch

* imple new proposer_shuffling_decision_root in callers

* push missed_blocks

* prune missed_blocks

* only add to hashmap if it's a monitored validator

* remove current_epoch dup + typos

* extract in func

* add prom metrics

* checkpoint is not only epoch but slot as well

* add safeguard if we start a new chain at slot 0

* clean

* remove unnecessary negative value for a slot

* typo in comment

* remove unused current_epoch

* share beacon_proposer_cache between validator_monitor and beacon_chain

* pass Hash256::zero()

* debug objects

* fix loop: lag is at the head

* sed s/get_slot/get_epoch

* fewer calls to cache.get_epoch

* fix typos

* remove cache first call

* export TYPICAL_SLOTS_PER_EPOCH and use it in validator_monitor

* switch to gauge & loop over missed_blocks hashset

* fix subnet_service tests

* remove unused var

* clean + fix nits

* add beacon_proposer_cache + validator_monitor in builder

* fix store_tests

* fix builder tests

* add tests

* add validator monitor set of tests

* clean tests

* nits

* optimise imports

* lint

* typo

* added self.aggregatable

* duplicate proposer_shuffling_decision_root

* remove duplication in passing beacon_proposer_cache

* remove duplication in passing beacon_proposer_cache

* using indices

* fmt

* implement missed blocks total

* nits

* avoid heap allocation

* remove recursion limit

* fix lint

* Fix valdiator monitor builder pattern

Unify validator monitor config struct

* renaming metrics

* renaming metrics in validator monitor

* add log if there's a missing validator index

* consistent log

* fix loop

* better loop

* move gauge to counter

* fmt

* add error message

* lint

* fix prom metrics

* set gauge to 0 when non-finalized epochs

* better wording

* remove hash256::zero in favour of block_root

* fix gauge total label

* fix last missed block validator

* Add `MissedBlock` struct

* Fix comment

* Refactor non-finalized block loop

* Fix off-by-one

* Avoid string allocation

* Fix compile error

* Remove non-finalized blocks metric

* fix func clojure

* remove unused variable

* remove unused DEFAULT_INDIVIDUAL_TRACKING_THRESHOLD

* remove unused DEFAULT_INDIVIDUAL_TRACKING_THRESHOLD in builder

* add validator index depending on the fork name

* typos

---------

Co-authored-by: Paul Hauner <paul@paulhauner.com>

beacon_node/client/src/builder.rs

@@ -191,15 +191,7 @@ where
             .graffiti(graffiti)
             .event_handler(event_handler)
             .execution_layer(execution_layer)
-            .monitor_validators(
-                config.validator_monitor_auto,
-                config.validator_monitor_pubkeys.clone(),
-                config.validator_monitor_individual_tracking_threshold,
-                runtime_context
-                    .service_context("val_mon".to_string())
-                    .log()
-                    .clone(),
-            );
+            .validator_monitor_config(config.validator_monitor.clone());
 
         let builder = if let Some(slasher) = self.slasher.clone() {
             builder.slasher(slasher)

beacon_node/client/src/config.rs

@@ -1,4 +1,4 @@
-use beacon_chain::validator_monitor::DEFAULT_INDIVIDUAL_TRACKING_THRESHOLD;
+use beacon_chain::validator_monitor::ValidatorMonitorConfig;
 use beacon_chain::TrustedSetup;
 use beacon_processor::BeaconProcessorConfig;
 use directory::DEFAULT_ROOT_DIR;
@@ -9,7 +9,7 @@ use serde::{Deserialize, Serialize};
 use std::fs;
 use std::path::PathBuf;
 use std::time::Duration;
-use types::{Graffiti, PublicKeyBytes};
+use types::Graffiti;
 /// Default directory name for the freezer database under the top-level data dir.
 const DEFAULT_FREEZER_DB_DIR: &str = "freezer_db";
 
@@ -56,15 +56,7 @@ pub struct Config {
     pub sync_eth1_chain: bool,
     /// Graffiti to be inserted everytime we create a block.
     pub graffiti: Graffiti,
-    /// When true, automatically monitor validators using the HTTP API.
-    pub validator_monitor_auto: bool,
-    /// A list of validator pubkeys to monitor.
-    pub validator_monitor_pubkeys: Vec<PublicKeyBytes>,
-    /// Once the number of monitored validators goes above this threshold, we
-    /// will stop tracking metrics on a per-validator basis. This prevents large
-    /// validator counts causing infeasibly high cardinailty for Prometheus and
-    /// high log volumes.
-    pub validator_monitor_individual_tracking_threshold: usize,
+    pub validator_monitor: ValidatorMonitorConfig,
     #[serde(skip)]
     /// The `genesis` field is not serialized or deserialized by `serde` to ensure it is defined
     /// via the CLI at runtime, instead of from a configuration file saved to disk.
@@ -107,9 +99,7 @@ impl Default for Config {
             http_metrics: <_>::default(),
             monitoring_api: None,
             slasher: None,
-            validator_monitor_auto: false,
-            validator_monitor_pubkeys: vec![],
-            validator_monitor_individual_tracking_threshold: DEFAULT_INDIVIDUAL_TRACKING_THRESHOLD,
+            validator_monitor: <_>::default(),
             logger_config: LoggerConfig::default(),
             beacon_processor: <_>::default(),
             genesis_state_url: <_>::default(),