tests: improve stability of test_deletion_queue_recovery (#7325)

## Problem

As https://github.com/neondatabase/neon/issues/6092 points out, this
test was (ab)using a failpoint!() with 'pause', which was occasionally
causing index uploads to get hung on a stuck executor thread, resulting
in timeouts waiting for remote_consistent_lsn.

That is one of several failure modes, but by far the most frequent.

## Summary of changes

- Replace the failpoint! with a `sleep_millis_async`, which is not only
async but also supports clean shutdown.
- Improve debugging: log the consistent LSN when scheduling an index
upload
- Tidy: remove an unnecessary checkpoint in the test code, where
last_flush_lsn_upload had just been called (this does a checkpoint
internally)

pageserver/src/control_plane_client.rs

@@ -12,7 +12,7 @@ use pageserver_api::{
 use serde::{de::DeserializeOwned, Serialize};
 use tokio_util::sync::CancellationToken;
 use url::Url;
-use utils::{backoff, generation::Generation, id::NodeId};
+use utils::{backoff, failpoint_support, generation::Generation, id::NodeId};
 
 use crate::{
     config::{NodeMetadata, PageServerConf},
@@ -210,7 +210,10 @@ impl ControlPlaneGenerationsApi for ControlPlaneClient {
                 .collect(),
         };
 
-        fail::fail_point!("control-plane-client-validate");
+        failpoint_support::sleep_millis_async!("control-plane-client-validate-sleep", &self.cancel);
+        if self.cancel.is_cancelled() {
+            return Err(RetryForeverError::ShuttingDown);
+        }
 
         let response: ValidateResponse = self.retry_http_forever(&re_attach_path, request).await?;
 

pageserver/src/tenant/remote_timeline_client.rs

@@ -593,14 +593,14 @@ impl RemoteTimelineClient {
         upload_queue: &mut UploadQueueInitialized,
         metadata: TimelineMetadata,
     ) {
+        let disk_consistent_lsn = upload_queue.latest_metadata.disk_consistent_lsn();
+
         info!(
-            "scheduling metadata upload with {} files ({} changed)",
+            "scheduling metadata upload up to consistent LSN {disk_consistent_lsn} with {} files ({} changed)",
             upload_queue.latest_files.len(),
             upload_queue.latest_files_changes_since_metadata_upload_scheduled,
         );
 
-        let disk_consistent_lsn = upload_queue.latest_metadata.disk_consistent_lsn();
-
         let index_part = IndexPart::new(
             upload_queue.latest_files.clone(),
             disk_consistent_lsn,

test_runner/regress/test_pageserver_generations.py

@@ -111,7 +111,6 @@ def generate_uploads_and_deletions(
             last_flush_lsn_upload(
                 env, endpoint, tenant_id, timeline_id, pageserver_id=pageserver.id
             )
-            ps_http.timeline_checkpoint(tenant_id, timeline_id)
 
         # Compaction should generate some GC-elegible layers
         for i in range(0, 2):
@@ -385,9 +384,8 @@ def test_deletion_queue_recovery(
     if validate_before == ValidateBefore.NO_VALIDATE:
         failpoints.append(
             # Prevent deletion lists from being validated, we will test that they are
-            # dropped properly during recovery.  'pause' is okay here because we kill
-            # the pageserver with immediate=true
-            ("control-plane-client-validate", "pause")
+            # dropped properly during recovery.  This is such a long sleep as to be equivalent to "never"
+            ("control-plane-client-validate", "return(3600000)")
         )
 
     ps_http.configure_failpoints(failpoints)