From a584e300d1ee4dd8ce6668f11e2720871fec3f7c Mon Sep 17 00:00:00 2001
From: Joonas Koivunen <joonas@neon.tech>
Date: Thu, 18 Jan 2024 12:39:45 +0200
Subject: [PATCH] test: figure out the relative eviction order assertions
 (#6375)

I just failed to see this earlier on #6136. layer counts are used as an
abstraction, and each of the two tenants lose proportionally about the
same amount of layers. sadly there is no difference in between
`relative_spare` and `relative_equal` as both of these end up evicting
the exact same amount of layers, but I'll try to add later another test
for those.

Cc: #5304
---
 .../regress/test_disk_usage_eviction.py       | 42 ++++++++++++++++++-
 1 file changed, 40 insertions(+), 2 deletions(-)

diff --git a/test_runner/regress/test_disk_usage_eviction.py b/test_runner/regress/test_disk_usage_eviction.py
index 64654c41a2..70c3b77516 100644
--- a/test_runner/regress/test_disk_usage_eviction.py
+++ b/test_runner/regress/test_disk_usage_eviction.py
@@ -1,5 +1,6 @@
 import enum
 import time
+from collections import Counter
 from dataclasses import dataclass
 from typing import Any, Dict, Tuple
 
@@ -119,6 +120,19 @@ class EvictionEnv:
             for tid, tlid in self.timelines
         }
 
+    def count_layers_per_tenant(self, pageserver: NeonPageserver) -> Dict[TenantId, int]:
+        ret: Counter[TenantId] = Counter()
+
+        for tenant_id, timeline_id in self.timelines:
+            timeline_dir = pageserver.timeline_dir(tenant_id, timeline_id)
+            assert timeline_dir.exists()
+            for file in timeline_dir.iterdir():
+                if "__" not in file.name:
+                    continue
+                ret[tenant_id] += 1
+
+        return dict(ret)
+
     def warm_up_tenant(self, tenant_id: TenantId):
         """
         Start a read-only compute at the LSN after pgbench -i, and run pgbench -S against it.
@@ -503,6 +517,7 @@ def test_partial_evict_tenant(eviction_env: EvictionEnv, order: EvictionOrder):
 
     (total_on_disk, _, _) = env.timelines_du(env.pageserver)
     du_by_timeline = env.du_by_timeline(env.pageserver)
+    tenant_layers = env.count_layers_per_tenant(env.pageserver)
 
     # pick smaller or greater (iteration order is insertion order of scale=4 and scale=6)
     [warm, cold] = list(du_by_timeline.keys())
@@ -556,8 +571,31 @@ def test_partial_evict_tenant(eviction_env: EvictionEnv, order: EvictionOrder):
             cold_size < cold_upper
         ), "the cold tenant should be evicted to its min_resident_size, i.e., max layer file size"
     else:
-        # just go with the space was freed, find proper limits later
-        pass
+        # with relative order what matters is the amount of layers, with a
+        # fudge factor of whether the eviction bothers tenants with highest
+        # layer count the most. last accessed times between tenants does not
+        # matter.
+        layers_now = env.count_layers_per_tenant(env.pageserver)
+
+        expected_ratio = later_total_on_disk / total_on_disk
+        log.info(
+            f"freed up {100 * expected_ratio}%, expecting the layer counts to decrease in similar ratio"
+        )
+
+        for tenant_id, original_count in tenant_layers.items():
+            count_now = layers_now[tenant_id]
+            ratio = count_now / original_count
+            abs_diff = abs(ratio - expected_ratio)
+            assert original_count > count_now
+            log.info(
+                f"tenant {tenant_id} layer count {original_count} -> {count_now}, ratio: {ratio}, expecting {abs_diff} < 0.1"
+            )
+
+            # in this test case both relative_spare and relative_equal produce
+            # the same outcomes; this must be a quantization effect of similar
+            # sizes (-s4 and -s6) and small (5MB) layer size.
+            # for pg15 and pg16 the absdiff is < 0.01, for pg14 it is closer to 0.02
+            assert abs_diff < 0.05
 
 
 def poor_mans_du(