make evictions_low_residence_duration_metric_threshold per-tenant (#3949)

Before this patch, if a tenant would override its eviction_policy
setting to use a lower LayerAccessThreshold::threshold than the
`evictions_low_residence_duration_metric_threshold`, the evictions done
for that tenant would count towards the
`evictions_with_low_residence_duration` metric.

That metric is used to identify pre-mature evictions, commonly triggered
by disk-usage-based eviction under disk pressure.

We don't want that to happen for the legitimate evictions of the tenant
that overrides its eviction_policy.

So, this patch
- moves the setting into TenantConf
- adds test coverage
- updates the staging & prod yamls

Forward Compatibility:
Software before this patch will ignore the new tenant conf field and use
the global one instead.
So we can roll back safely.

Backward Compatibility:
Parsing old configs with software as of this patch will fail in
`PageServerConf::parse_and_validate` with error 
`unrecognized pageserver option 'evictions_low_residence_duration_metric_threshold'`
if the option is still present in the global section.
We deal with this by updating the configs in Ansible.

fixes https://github.com/neondatabase/neon/issues/3940

test_runner/fixtures/pageserver/http.py

@@ -519,6 +519,13 @@ class PageserverHttpClient(requests.Session):
 
         assert res.status_code == 200
 
+    def download_all_layers(self, tenant_id: TenantId, timeline_id: TimelineId):
+        info = self.layer_map_info(tenant_id, timeline_id)
+        for layer in info.historic_layers:
+            if not layer.remote:
+                continue
+            self.download_layer(tenant_id, timeline_id, layer.layer_file_name)
+
     def evict_layer(self, tenant_id: TenantId, timeline_id: TimelineId, layer_name: str):
         res = self.delete(
             f"http://localhost:{self.port}/v1/tenant/{tenant_id}/timeline/{timeline_id}/layer/{layer_name}",

test_runner/regress/test_tenant_conf.py

@@ -18,7 +18,11 @@ def test_tenant_config(neon_env_builder: NeonEnvBuilder):
     neon_env_builder.pageserver_config_override = """
 page_cache_size=444;
 wait_lsn_timeout='111 s';
-tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
+[tenant_config]
+checkpoint_distance = 10000
+compaction_target_size = 1048576
+evictions_low_residence_duration_metric_threshold = "2 days"
+"""
 
     env = neon_env_builder.init_start()
     http_client = env.pageserver.http_client()
@@ -39,6 +43,7 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     new_conf = {
         "checkpoint_distance": "20000",
         "gc_period": "30sec",
+        "evictions_low_residence_duration_metric_threshold": "42s",
     }
     tenant, _ = env.neon_cli.create_tenant(conf=new_conf)
 
@@ -78,6 +83,7 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     assert effective_config["gc_period"] == "1h"
     assert effective_config["image_creation_threshold"] == 3
     assert effective_config["pitr_interval"] == "7days"
+    assert effective_config["evictions_low_residence_duration_metric_threshold"] == "2days"
 
     # check the configuration of the new tenant
     with closing(env.pageserver.connect()) as psconn:
@@ -112,6 +118,9 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     assert (
         new_effective_config["gc_period"] == "30s"
     ), "Specific 'gc_period' config should override the default value"
+    assert (
+        new_effective_config["evictions_low_residence_duration_metric_threshold"] == "42s"
+    ), "Should override default value"
     assert new_effective_config["compaction_target_size"] == 1048576
     assert new_effective_config["compaction_period"] == "20s"
     assert new_effective_config["compaction_threshold"] == 10
@@ -125,6 +134,7 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
         "gc_period": "80sec",
         "compaction_period": "80sec",
         "image_creation_threshold": "2",
+        "evictions_low_residence_duration_metric_threshold": "23h",
     }
     env.neon_cli.config_tenant(
         tenant_id=tenant,
@@ -167,6 +177,9 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     assert (
         updated_effective_config["compaction_period"] == "1m 20s"
     ), "Specific 'compaction_period' config should override the default value"
+    assert (
+        updated_effective_config["evictions_low_residence_duration_metric_threshold"] == "23h"
+    ), "Should override default value"
     assert updated_effective_config["compaction_target_size"] == 1048576
     assert updated_effective_config["compaction_threshold"] == 10
     assert updated_effective_config["gc_horizon"] == 67108864
@@ -225,6 +238,7 @@ tenant_config={checkpoint_distance = 10000, compaction_target_size = 1048576}"""
     assert final_effective_config["gc_horizon"] == 67108864
     assert final_effective_config["gc_period"] == "1h"
     assert final_effective_config["image_creation_threshold"] == 3
+    assert final_effective_config["evictions_low_residence_duration_metric_threshold"] == "2days"
 
     # restart the pageserver and ensure that the config is still correct
     env.pageserver.stop()
@@ -285,3 +299,81 @@ def test_creating_tenant_conf_after_attach(neon_env_builder: NeonEnvBuilder):
     # dont test applying the setting here, we have that another test case to show it
     # we just care about being able to create the file
     assert len(contents_first) > len(contents_later)
+
+
+def test_live_reconfig_get_evictions_low_residence_duration_metric_threshold(
+    neon_env_builder: NeonEnvBuilder,
+):
+    neon_env_builder.enable_remote_storage(
+        remote_storage_kind=RemoteStorageKind.LOCAL_FS,
+        test_name="test_live_reconfig_get_evictions_low_residence_duration_metric_threshold",
+    )
+
+    env = neon_env_builder.init_start()
+    assert isinstance(env.remote_storage, LocalFsStorage)
+
+    (tenant_id, timeline_id) = env.neon_cli.create_tenant()
+    ps_http = env.pageserver.http_client()
+
+    def get_metric():
+        metrics = ps_http.get_metrics()
+        metric = metrics.query_one(
+            "pageserver_evictions_with_low_residence_duration_total",
+            {
+                "tenant_id": str(tenant_id),
+                "timeline_id": str(timeline_id),
+            },
+        )
+        return metric
+
+    default_value = ps_http.tenant_config(tenant_id).effective_config[
+        "evictions_low_residence_duration_metric_threshold"
+    ]
+    metric = get_metric()
+    assert int(metric.value) == 0, "metric is present with default value"
+
+    assert default_value == "1day"
+
+    ps_http.download_all_layers(tenant_id, timeline_id)
+    ps_http.evict_all_layers(tenant_id, timeline_id)
+    metric = get_metric()
+    assert int(metric.value) > 0, "metric is updated"
+
+    env.neon_cli.config_tenant(
+        tenant_id, {"evictions_low_residence_duration_metric_threshold": default_value}
+    )
+    updated_metric = get_metric()
+    assert int(updated_metric.value) == int(
+        metric.value
+    ), "metric is unchanged when setting same value"
+
+    env.neon_cli.config_tenant(
+        tenant_id, {"evictions_low_residence_duration_metric_threshold": "2day"}
+    )
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 2 * 24 * 60 * 60
+    assert int(metric.value) == 0
+
+    ps_http.download_all_layers(tenant_id, timeline_id)
+    ps_http.evict_all_layers(tenant_id, timeline_id)
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 2 * 24 * 60 * 60
+    assert int(metric.value) > 0
+
+    env.neon_cli.config_tenant(
+        tenant_id, {"evictions_low_residence_duration_metric_threshold": "2h"}
+    )
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 2 * 60 * 60
+    assert int(metric.value) == 0, "value resets if label changes"
+
+    ps_http.download_all_layers(tenant_id, timeline_id)
+    ps_http.evict_all_layers(tenant_id, timeline_id)
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 2 * 60 * 60
+    assert int(metric.value) > 0, "set a non-zero value for next step"
+
+    env.neon_cli.config_tenant(tenant_id, {})
+    metric = get_metric()
+    assert int(metric.labels["low_threshold_secs"]) == 24 * 60 * 60, "label resets to default"
+    assert int(metric.value) == 0, "value resets to default"