pageserver: layer count & size metrics (#8410)

## Problem

We lack insight into:
- How much of a tenant's physical size is image vs. delta layers
- Average sizes of image vs. delta layers
- Total layer counts per timeline, indicating size of index_part object

As well as general observability love, this is motivated by
https://github.com/neondatabase/neon/issues/6738, where we need to
define some sensible thresholds for storage amplification, and using
total physical size may not work well (if someone does a lot of DROPs
then it's legitimate for the physical-synthetic ratio to be huge), but
the ratio between image layer size and delta layer size may be a better
indicator of whether we're generating unreasonable quantities of image
layers.

## Summary of changes

- Add pageserver_layer_bytes and pageserver_layer_count metrics,
labelled by timeline and `kind` (delta or image)
- Add & subtract these with LayerInner's lifetime.

I'm intentionally avoiding using a generic metric RAII guard object, to
avoid bloating LayerInner: it already has all the information it needs
to update metric on new+drop.

pageserver/src/metrics.rs

@@ -473,6 +473,31 @@ static PITR_HISTORY_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
+#[derive(strum_macros::EnumString, strum_macros::Display, strum_macros::IntoStaticStr)]
+#[strum(serialize_all = "kebab_case")]
+pub(crate) enum MetricLayerKind {
+    Delta,
+    Image,
+}
+
+static TIMELINE_LAYER_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
+    register_uint_gauge_vec!(
+        "pageserver_layer_bytes",
+        "Sum of layer physical sizes in bytes",
+        &["tenant_id", "shard_id", "timeline_id", "kind"]
+    )
+    .expect("failed to define a metric")
+});
+
+static TIMELINE_LAYER_COUNT: Lazy<UIntGaugeVec> = Lazy::new(|| {
+    register_uint_gauge_vec!(
+        "pageserver_layer_count",
+        "Number of layers that exist",
+        &["tenant_id", "shard_id", "timeline_id", "kind"]
+    )
+    .expect("failed to define a metric")
+});
+
 static TIMELINE_ARCHIVE_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
     register_uint_gauge_vec!(
         "pageserver_archive_size",
@@ -2141,6 +2166,10 @@ pub(crate) struct TimelineMetrics {
     pub last_record_gauge: IntGauge,
     pub pitr_history_size: UIntGauge,
     pub archival_size: UIntGauge,
+    pub(crate) layer_size_image: UIntGauge,
+    pub(crate) layer_count_image: UIntGauge,
+    pub(crate) layer_size_delta: UIntGauge,
+    pub(crate) layer_count_delta: UIntGauge,
     pub standby_horizon_gauge: IntGauge,
     pub resident_physical_size_gauge: UIntGauge,
     /// copy of LayeredTimeline.current_logical_size
@@ -2223,6 +2252,42 @@ impl TimelineMetrics {
             .get_metric_with_label_values(&[&tenant_id, &shard_id, &timeline_id])
             .unwrap();
 
+        let layer_size_image = TIMELINE_LAYER_SIZE
+            .get_metric_with_label_values(&[
+                &tenant_id,
+                &shard_id,
+                &timeline_id,
+                MetricLayerKind::Image.into(),
+            ])
+            .unwrap();
+
+        let layer_count_image = TIMELINE_LAYER_COUNT
+            .get_metric_with_label_values(&[
+                &tenant_id,
+                &shard_id,
+                &timeline_id,
+                MetricLayerKind::Image.into(),
+            ])
+            .unwrap();
+
+        let layer_size_delta = TIMELINE_LAYER_SIZE
+            .get_metric_with_label_values(&[
+                &tenant_id,
+                &shard_id,
+                &timeline_id,
+                MetricLayerKind::Delta.into(),
+            ])
+            .unwrap();
+
+        let layer_count_delta = TIMELINE_LAYER_COUNT
+            .get_metric_with_label_values(&[
+                &tenant_id,
+                &shard_id,
+                &timeline_id,
+                MetricLayerKind::Delta.into(),
+            ])
+            .unwrap();
+
         let standby_horizon_gauge = STANDBY_HORIZON
             .get_metric_with_label_values(&[&tenant_id, &shard_id, &timeline_id])
             .unwrap();
@@ -2277,6 +2342,10 @@ impl TimelineMetrics {
             last_record_gauge,
             pitr_history_size,
             archival_size,
+            layer_size_image,
+            layer_count_image,
+            layer_size_delta,
+            layer_count_delta,
             standby_horizon_gauge,
             resident_physical_size_gauge,
             current_logical_size_gauge,
@@ -2338,6 +2407,31 @@ impl TimelineMetrics {
         let _ = TIMELINE_ARCHIVE_SIZE.remove_label_values(&[tenant_id, shard_id, timeline_id]);
         let _ = PITR_HISTORY_SIZE.remove_label_values(&[tenant_id, shard_id, timeline_id]);
 
+        let _ = TIMELINE_LAYER_SIZE.remove_label_values(&[
+            tenant_id,
+            shard_id,
+            timeline_id,
+            MetricLayerKind::Image.into(),
+        ]);
+        let _ = TIMELINE_LAYER_COUNT.remove_label_values(&[
+            tenant_id,
+            shard_id,
+            timeline_id,
+            MetricLayerKind::Image.into(),
+        ]);
+        let _ = TIMELINE_LAYER_SIZE.remove_label_values(&[
+            tenant_id,
+            shard_id,
+            timeline_id,
+            MetricLayerKind::Delta.into(),
+        ]);
+        let _ = TIMELINE_LAYER_COUNT.remove_label_values(&[
+            tenant_id,
+            shard_id,
+            timeline_id,
+            MetricLayerKind::Delta.into(),
+        ]);
+
         let _ = EVICTIONS.remove_label_values(&[tenant_id, shard_id, timeline_id]);
         let _ = AUX_FILE_SIZE.remove_label_values(&[tenant_id, shard_id, timeline_id]);
         let _ = VALID_LSN_LEASE_COUNT.remove_label_values(&[tenant_id, shard_id, timeline_id]);

pageserver/src/tenant/storage_layer/layer.rs

@@ -693,6 +693,18 @@ impl Drop for LayerInner {
             // and we could be delaying shutdown for nothing.
         }
 
+        if let Some(timeline) = self.timeline.upgrade() {
+            // Only need to decrement metrics if the timeline still exists: otherwise
+            // it will have already de-registered these metrics via TimelineMetrics::shutdown
+            if self.desc.is_delta() {
+                timeline.metrics.layer_count_delta.dec();
+                timeline.metrics.layer_size_delta.sub(self.desc.file_size);
+            } else {
+                timeline.metrics.layer_count_image.dec();
+                timeline.metrics.layer_size_image.sub(self.desc.file_size);
+            }
+        }
+
         if !*self.wanted_deleted.get_mut() {
             return;
         }
@@ -791,6 +803,15 @@ impl LayerInner {
             (heavier_once_cell::OnceCell::default(), 0, Status::Evicted)
         };
 
+        // This object acts as a RAII guard on these metrics: increment on construction
+        if desc.is_delta() {
+            timeline.metrics.layer_count_delta.inc();
+            timeline.metrics.layer_size_delta.add(desc.file_size);
+        } else {
+            timeline.metrics.layer_count_image.inc();
+            timeline.metrics.layer_size_image.add(desc.file_size);
+        }
+
         LayerInner {
             conf,
             debug_str: {

test_runner/fixtures/metrics.py

@@ -146,6 +146,8 @@ PAGESERVER_PER_TENANT_METRICS: Tuple[str, ...] = (
     "pageserver_smgr_query_seconds_sum",
     "pageserver_archive_size",
     "pageserver_pitr_history_size",
+    "pageserver_layer_bytes",
+    "pageserver_layer_count",
     "pageserver_storage_operations_seconds_count_total",
     "pageserver_storage_operations_seconds_sum_total",
     "pageserver_evictions_total",