Add config value MetatdataMonitoringEnabledjc-metadata-monitoring-flag

With multiple Praefect instances running, each one will hit the database
for metrics. This causes unnecessary load on the database and in some
cases, fatally so. To fix this, we will add a new config value so that
we can control whicih Praefect serves as thei one that will monitor the
database.

Changelog: added

2 files changed, 3 insertions, 1 deletions

diff --git a/cmd/praefect/main.go b/cmd/praefect/main.go
index e2210e659..40e45b370 100644
--- a/cmd/praefect/main.go
+++ b/cmd/praefect/main.go
@@ -390,7 +390,7 @@ func run(cfgs []starter.Config, conf config.Config, b bootstrap.Listener, promre
 		)
 	)
 	metricsCollectors = append(metricsCollectors, transactionManager, coordinator, repl)
-	if db != nil {
+	if db != nil && conf.Prometheus.MonitorMetadata {
 		promreg.MustRegister(
 			datastore.NewRepositoryStoreCollector(logger, conf.VirtualStorageNames(), db, conf.Prometheus.ScrapeTimeout),
 		)
diff --git a/internal/gitaly/config/prometheus/config.go b/internal/gitaly/config/prometheus/config.go
index 629eca6a5..8649ab560 100644
--- a/internal/gitaly/config/prometheus/config.go
+++ b/internal/gitaly/config/prometheus/config.go
@@ -16,6 +16,8 @@ type Config struct {
 	// GRPCLatencyBuckets configures the histogram buckets used for gRPC
 	// latency measurements.
 	GRPCLatencyBuckets []float64 `toml:"grpc_latency_buckets"`
+	// MonitorMetadata when true, will register collectors that query the praefect database
+	MonitorMetadata bool `toml:"monitor_metadata"`
 }
 
 // DefaultConfig returns a new config with default values set.