feat(pageserver): add an optional lease to the get_lsn_by_timestamp API

pageserver/src/http/openapi_spec.yml

@@ -251,6 +251,13 @@ paths:
             type: string
             format: date-time
           description: A timestamp to get the LSN
+        - name: with_lease
+          in: query
+          required: false
+          schema:
+            type: boolean
+          description: Whether to grant a lease to the corresponding LSN. Default to false.
+
       responses:
         "200":
           description: OK
@@ -1086,6 +1093,10 @@ components:
         kind:
           type: string
           enum: [past, present, future, nodata]
+        valid_until:
+          type: string
+          format: date-time
+          description: The expiration time of the granted lease.
 
     LsnLease:
       type: object

pageserver/src/http/routes.rs

@@ -21,6 +21,7 @@ use pageserver_api::models::IngestAuxFilesRequest;
 use pageserver_api::models::ListAuxFilesRequest;
 use pageserver_api::models::LocationConfig;
 use pageserver_api::models::LocationConfigListResponse;
+use pageserver_api::models::LsnLease;
 use pageserver_api::models::ShardParameters;
 use pageserver_api::models::TenantDetails;
 use pageserver_api::models::TenantLocationConfigResponse;
@@ -735,6 +736,8 @@ async fn get_lsn_by_timestamp_handler(
         .map_err(ApiError::BadRequest)?;
     let timestamp_pg = postgres_ffi::to_pg_timestamp(timestamp);
 
+    let with_lease = parse_query_param(&request, "with_lease")?.unwrap_or(false);
+
     let ctx = RequestContext::new(TaskKind::MgmtRequest, DownloadBehavior::Download);
 
     let timeline =
@@ -743,22 +746,41 @@ async fn get_lsn_by_timestamp_handler(
     let result = timeline
         .find_lsn_for_timestamp(timestamp_pg, &cancel, &ctx)
         .await?;
+
     #[derive(serde::Serialize, Debug)]
     struct Result {
         lsn: Lsn,
         kind: &'static str,
+        #[serde(default)]
+        #[serde(skip_serializing_if = "Option::is_none")]
+        #[serde(flatten)]
+        lease: Option<LsnLease>,
     }
     let (lsn, kind) = match result {
         LsnForTimestamp::Present(lsn) => (lsn, "present"),
         LsnForTimestamp::Future(lsn) => (lsn, "future"),
         LsnForTimestamp::Past(lsn) => (lsn, "past"),
         LsnForTimestamp::NoData(lsn) => (lsn, "nodata"),
     };
-    let result = Result { lsn, kind };
+
+    let lease = if with_lease && (kind == "present" || kind == "future") {
+        timeline
+            .make_lsn_lease(lsn, timeline.get_lsn_lease_length_for_ts(), &ctx)
+            .ok()
+    } else {
+        None
+    };
+
+    let result = Result { lsn, kind, lease };
+    let valid_until = result
+        .lease
+        .as_ref()
+        .map(|l| humantime::format_rfc3339_millis(l.valid_until).to_string());
     tracing::info!(
         lsn=?result.lsn,
         kind=%result.kind,
         timestamp=%timestamp_raw,
+        valid_until=?valid_until,
         "lsn_by_timestamp finished"
     );
     json_response(StatusCode::OK, result)

test_runner/fixtures/pageserver/http.py

@@ -630,13 +630,15 @@ def timeline_get_lsn_by_timestamp(
         tenant_id: Union[TenantId, TenantShardId],
         timeline_id: TimelineId,
         timestamp: datetime,
+        with_lease: bool = False,
         **kwargs,
     ):
         log.info(
-            f"Requesting lsn by timestamp {timestamp}, tenant {tenant_id}, timeline {timeline_id}"
+            f"Requesting lsn by timestamp {timestamp}, tenant {tenant_id}, timeline {timeline_id}, with_lease={with_lease}"
         )
+        with_lease_query = f"with_lease={with_lease}".lower()
         res = self.get(
-            f"http://localhost:{self.port}/v1/tenant/{tenant_id}/timeline/{timeline_id}/get_lsn_by_timestamp?timestamp={timestamp.isoformat()}Z",
+            f"http://localhost:{self.port}/v1/tenant/{tenant_id}/timeline/{timeline_id}/get_lsn_by_timestamp?timestamp={timestamp.isoformat()}Z&{with_lease_query}",
             **kwargs,
         )
         self.verbose_error(res)

test_runner/regress/test_lsn_mapping.py

@@ -12,10 +12,29 @@
 from requests.exceptions import ReadTimeout
 
 
-#
-# Test pageserver get_lsn_by_timestamp API
-#
-def test_lsn_mapping(neon_env_builder: NeonEnvBuilder):
+def assert_lsn_lease_granted(result, with_lease: bool):
+    """
+    Asserts an LSN lease is granted when `with_lease` flag is turned on.
+    Always asserts no LSN lease is granted when `with_lease` flag is off.
+    """
+    if with_lease:
+        assert result.get("valid_until")
+    else:
+        assert result.get("valid_until") is None
+
+
+def assert_lsn_lease_not_granted(result):
+    """
+    Asserts no LSN lease is not granted.
+    """
+    assert result.get("valid_until") is None
+
+
+@pytest.mark.parametrize("with_lease", [True, False])
+def test_lsn_mapping(neon_env_builder: NeonEnvBuilder, with_lease: bool):
+    """
+    Test pageserver get_lsn_by_timestamp API.
+    """
     env = neon_env_builder.init_start()
 
     tenant_id, _ = env.neon_cli.create_tenant(
@@ -67,23 +86,33 @@ def test_lsn_mapping(neon_env_builder: NeonEnvBuilder):
         # Check edge cases
         # Timestamp is in the future
         probe_timestamp = tbl[-1][1] + timedelta(hours=1)
-        result = client.timeline_get_lsn_by_timestamp(tenant_id, timeline_id, probe_timestamp)
+        result = client.timeline_get_lsn_by_timestamp(
+            tenant_id, timeline_id, probe_timestamp, with_lease=with_lease
+        )
         assert result["kind"] == "future"
+        assert_lsn_lease_granted(result, with_lease)
         # make sure that we return a well advanced lsn here
         assert Lsn(result["lsn"]) > start_lsn
 
         # Timestamp is in the unreachable past
         probe_timestamp = tbl[0][1] - timedelta(hours=10)
-        result = client.timeline_get_lsn_by_timestamp(tenant_id, timeline_id, probe_timestamp)
+        result = client.timeline_get_lsn_by_timestamp(
+            tenant_id, timeline_id, probe_timestamp, with_lease=with_lease
+        )
         assert result["kind"] == "past"
+        assert_lsn_lease_not_granted(result)
+
         # make sure that we return the minimum lsn here at the start of the range
         assert Lsn(result["lsn"]) < start_lsn
 
         # Probe a bunch of timestamps in the valid range
         for i in range(1, len(tbl), 100):
             probe_timestamp = tbl[i][1]
-            result = client.timeline_get_lsn_by_timestamp(tenant_id, timeline_id, probe_timestamp)
+            result = client.timeline_get_lsn_by_timestamp(
+                tenant_id, timeline_id, probe_timestamp, with_lease=with_lease
+            )
             assert result["kind"] not in ["past", "nodata"]
+            assert_lsn_lease_granted(result, with_lease)
             lsn = result["lsn"]
             # Call get_lsn_by_timestamp to get the LSN
             # Launch a new read-only node at that LSN, and check that only the rows
@@ -105,8 +134,11 @@ def test_lsn_mapping(neon_env_builder: NeonEnvBuilder):
 
         # Timestamp is in the unreachable past
         probe_timestamp = tbl[0][1] - timedelta(hours=10)
-        result = client.timeline_get_lsn_by_timestamp(tenant_id, timeline_id_child, probe_timestamp)
+        result = client.timeline_get_lsn_by_timestamp(
+            tenant_id, timeline_id_child, probe_timestamp, with_lease=with_lease
+        )
         assert result["kind"] == "past"
+        assert_lsn_lease_not_granted(result)
         # make sure that we return the minimum lsn here at the start of the range
         assert Lsn(result["lsn"]) >= last_flush_lsn