From beefc7a8108e5af333bc1e453749acf872f18fdd Mon Sep 17 00:00:00 2001
From: John Spray <john@neon.tech>
Date: Tue, 20 Aug 2024 19:47:42 +0100
Subject: [PATCH] pageserver: add metric
 pageserver_secondary_heatmap_total_size (#8768)

## Problem

We don't have a convenient way for a human to ask "how far are secondary
downloads along for this tenant".

This is useful when driving migrations of tenants to the storage
controller, as we first create a secondary location and want to see it
warm up before we cut over. That can already be done via storcon_cli,
but we would like a way that doesn't require direct API access.

## Summary of changes

Add a metric that reports to total size of layers in the heatmap: this
may be used in conjunction with the existing
`pageserver_secondary_resident_physical_size` to estimate "warmth" of
the secondary location.
---
 pageserver/src/metrics.rs                     |  9 +++++++++
 pageserver/src/tenant/secondary.rs            | 10 ++++++++++
 pageserver/src/tenant/secondary/downloader.rs |  6 ++++++
 3 files changed, 25 insertions(+)

diff --git a/pageserver/src/metrics.rs b/pageserver/src/metrics.rs
index cd2cd43f27..1bc9352256 100644
--- a/pageserver/src/metrics.rs
+++ b/pageserver/src/metrics.rs
@@ -1803,6 +1803,15 @@ pub(crate) static SECONDARY_RESIDENT_PHYSICAL_SIZE: Lazy<UIntGaugeVec> = Lazy::n
     .expect("failed to define a metric")
 });
 
+pub(crate) static SECONDARY_HEATMAP_TOTAL_SIZE: Lazy<UIntGaugeVec> = Lazy::new(|| {
+    register_uint_gauge_vec!(
+        "pageserver_secondary_heatmap_total_size",
+        "The total size in bytes of all layers in the most recently downloaded heatmap.",
+        &["tenant_id", "shard_id"]
+    )
+    .expect("failed to define a metric")
+});
+
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 pub enum RemoteOpKind {
     Upload,
diff --git a/pageserver/src/tenant/secondary.rs b/pageserver/src/tenant/secondary.rs
index 3132a28b12..1331c07d05 100644
--- a/pageserver/src/tenant/secondary.rs
+++ b/pageserver/src/tenant/secondary.rs
@@ -8,6 +8,7 @@ use std::{sync::Arc, time::SystemTime};
 use crate::{
     context::RequestContext,
     disk_usage_eviction_task::DiskUsageEvictionInfo,
+    metrics::SECONDARY_HEATMAP_TOTAL_SIZE,
     task_mgr::{self, TaskKind, BACKGROUND_RUNTIME},
 };
 
@@ -105,6 +106,9 @@ pub(crate) struct SecondaryTenant {
 
     // Sum of layer sizes on local disk
     pub(super) resident_size_metric: UIntGauge,
+
+    // Sum of layer sizes in the most recently downloaded heatmap
+    pub(super) heatmap_total_size_metric: UIntGauge,
 }
 
 impl Drop for SecondaryTenant {
@@ -112,6 +116,7 @@ impl Drop for SecondaryTenant {
         let tenant_id = self.tenant_shard_id.tenant_id.to_string();
         let shard_id = format!("{}", self.tenant_shard_id.shard_slug());
         let _ = SECONDARY_RESIDENT_PHYSICAL_SIZE.remove_label_values(&[&tenant_id, &shard_id]);
+        let _ = SECONDARY_HEATMAP_TOTAL_SIZE.remove_label_values(&[&tenant_id, &shard_id]);
     }
 }
 
@@ -128,6 +133,10 @@ impl SecondaryTenant {
             .get_metric_with_label_values(&[&tenant_id, &shard_id])
             .unwrap();
 
+        let heatmap_total_size_metric = SECONDARY_HEATMAP_TOTAL_SIZE
+            .get_metric_with_label_values(&[&tenant_id, &shard_id])
+            .unwrap();
+
         Arc::new(Self {
             tenant_shard_id,
             // todo: shall we make this a descendent of the
@@ -145,6 +154,7 @@ impl SecondaryTenant {
             progress: std::sync::Mutex::default(),
 
             resident_size_metric,
+            heatmap_total_size_metric,
         })
     }
 
diff --git a/pageserver/src/tenant/secondary/downloader.rs b/pageserver/src/tenant/secondary/downloader.rs
index 8cff1d2864..90e1c01dbd 100644
--- a/pageserver/src/tenant/secondary/downloader.rs
+++ b/pageserver/src/tenant/secondary/downloader.rs
@@ -829,6 +829,12 @@ impl<'a> TenantDownloader<'a> {
             layers_downloaded: 0,
             bytes_downloaded: 0,
         };
+
+        // Also expose heatmap bytes_total as a metric
+        self.secondary_state
+            .heatmap_total_size_metric
+            .set(heatmap_stats.bytes);
+
         // Accumulate list of things to delete while holding the detail lock, for execution after dropping the lock
         let mut delete_layers = Vec::new();
         let mut delete_timelines = Vec::new();