Generate pgbench data on the server for most tests

This should generally be faster when running tests, especially those
that run with higher scales.

Ignoring test_lfc_resize since it seems like we are hitting a query
timeout for some reason that I have yet to investigate. A little bit of
improvemnt is better than none.

Signed-off-by: Tristan Partin <tristan@neon.tech>

test_runner/performance/test_branch_creation.py

@@ -107,7 +107,7 @@ def test_branch_creation_many(neon_compare: NeonCompare, n_branches: int, shape:
     env.neon_cli.create_branch("b0")
 
     endpoint = env.endpoints.create_start("b0")
-    neon_compare.pg_bin.run_capture(["pgbench", "-i", "-s10", endpoint.connstr()])
+    neon_compare.pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s10", endpoint.connstr()])
 
     branch_creation_durations = []
 

test_runner/performance/test_branching.py

@@ -43,7 +43,7 @@ def test_compare_child_and_root_pgbench_perf(neon_compare: NeonCompare):
 
     env.neon_cli.create_branch("root")
     endpoint_root = env.endpoints.create_start("root")
-    pg_bin.run_capture(["pgbench", "-i", endpoint_root.connstr(), "-s10"])
+    pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", endpoint_root.connstr(), "-s10"])
 
     fork_at_current_lsn(env, endpoint_root, "child", "root")
 

test_runner/performance/test_logical_replication.py

@@ -24,13 +24,13 @@ def test_logical_replication(neon_simple_env: NeonEnv, pg_bin: PgBin, vanilla_pg
 
     endpoint = env.endpoints.create_start("main")
 
-    pg_bin.run_capture(["pgbench", "-i", "-s10", endpoint.connstr()])
+    pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s10", endpoint.connstr()])
 
     endpoint.safe_psql("create publication pub1 for table pgbench_accounts, pgbench_history")
 
     # now start subscriber
     vanilla_pg.start()
-    pg_bin.run_capture(["pgbench", "-i", "-s10", vanilla_pg.connstr()])
+    pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s10", vanilla_pg.connstr()])
 
     vanilla_pg.safe_psql("truncate table pgbench_accounts")
     vanilla_pg.safe_psql("truncate table pgbench_history")
@@ -99,9 +99,9 @@ def test_subscriber_lag(
     sub_connstr = benchmark_project_sub.connstr
 
     if benchmark_project_pub.is_new:
-        pg_bin.run_capture(["pgbench", "-i", "-s100"], env=pub_env)
+        pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s100"], env=pub_env)
     if benchmark_project_sub.is_new:
-        pg_bin.run_capture(["pgbench", "-i", "-s100"], env=sub_env)
+        pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s100"], env=sub_env)
 
     pub_conn = psycopg2.connect(pub_connstr)
     sub_conn = psycopg2.connect(sub_connstr)
@@ -193,8 +193,8 @@ def test_publisher_restart(
     pub_connstr = benchmark_project_pub.connstr
     sub_connstr = benchmark_project_sub.connstr
 
-    pg_bin.run_capture(["pgbench", "-i", "-s100"], env=pub_env)
-    pg_bin.run_capture(["pgbench", "-i", "-s100"], env=sub_env)
+    pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s100"], env=pub_env)
+    pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s100"], env=sub_env)
 
     pub_conn = psycopg2.connect(pub_connstr)
     sub_conn = psycopg2.connect(sub_connstr)
@@ -288,7 +288,7 @@ def test_snap_files(
             is_super = cur.fetchall()[0][0]
             assert is_super, "This benchmark won't work if we don't have superuser"
 
-    pg_bin.run_capture(["pgbench", "-i", "-s100"], env=env)
+    pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s100"], env=env)
 
     conn = psycopg2.connect(connstr)
     conn.autocommit = True

test_runner/performance/test_physical_replication.py

@@ -85,7 +85,7 @@ def test_ro_replica_lag(
             endpoint_id=replica["endpoint"]["id"],
         )["uri"]
 
-        pg_bin.run_capture(["pgbench", "-i", "-s100"], env=master_env)
+        pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s100"], env=master_env)
 
         master_workload = pg_bin.run_nonblocking(
             ["pgbench", "-c10", pgbench_duration, "-Mprepared"],
@@ -212,7 +212,7 @@ def test_replication_start_stop(
         for i in range(num_replicas):
             replica_env[i]["PGHOST"] = replicas[i]["endpoint"]["host"]
 
-        pg_bin.run_capture(["pgbench", "-i", "-s10"], env=master_env)
+        pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s10"], env=master_env)
 
         # Sync replicas
         with psycopg2.connect(master_connstr) as conn_master:

test_runner/regress/test_branching.py

@@ -52,7 +52,7 @@ def test_branching_with_pgbench(
     def run_pgbench(connstr: str):
         log.info(f"Start a pgbench workload on pg {connstr}")
 
-        pg_bin.run_capture(["pgbench", "-i", f"-s{scale}", connstr])
+        pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", f"-s{scale}", connstr])
         pg_bin.run_capture(["pgbench", "-T15", connstr])
 
     env.neon_cli.create_branch("b0", tenant_id=tenant)

test_runner/regress/test_disk_usage_eviction.py

@@ -291,7 +291,7 @@ def pgbench_init_tenant(
     )
 
     with env.endpoints.create_start("main", tenant_id=tenant_id) as endpoint:
-        pg_bin.run(["pgbench", "-i", f"-s{scale}", endpoint.connstr()])
+        pg_bin.run(["pgbench", "-i", "-I", "dtGvp", f"-s{scale}", endpoint.connstr()])
         wait_for_last_flush_lsn(env, endpoint, tenant_id, timeline_id)
 
     return (tenant_id, timeline_id)

test_runner/regress/test_hot_standby.py

@@ -199,7 +199,7 @@ def test_hot_standby_gc(neon_env_builder: NeonEnvBuilder, pause_apply: bool):
 def run_pgbench(connstr: str, pg_bin: PgBin):
     log.info(f"Start a pgbench workload on pg {connstr}")
     # s10 is about 150MB of data. In debug mode init takes about 15s on SSD.
-    pg_bin.run_capture(["pgbench", "-i", "-s10", connstr])
+    pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", "-s10", connstr])
     log.info("pgbench init done")
     pg_bin.run_capture(["pgbench", "-T60", connstr])
 

test_runner/regress/test_pageserver_reconnect.py

@@ -22,7 +22,7 @@ def test_pageserver_reconnect(neon_simple_env: NeonEnv, pg_bin: PgBin):
 
     def run_pgbench(connstr: str):
         log.info(f"Start a pgbench workload on pg {connstr}")
-        pg_bin.run_capture(["pgbench", "-i", f"-s{scale}", connstr])
+        pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", f"-s{scale}", connstr])
         pg_bin.run_capture(["pgbench", f"-T{int(n_reconnects*timeout)}", connstr])
 
     thread = threading.Thread(target=run_pgbench, args=(endpoint.connstr(),), daemon=True)

test_runner/regress/test_pageserver_restarts_under_workload.py

@@ -19,7 +19,7 @@ def test_pageserver_restarts_under_worload(neon_simple_env: NeonEnv, pg_bin: PgB
 
     def run_pgbench(connstr: str):
         log.info(f"Start a pgbench workload on pg {connstr}")
-        pg_bin.run_capture(["pgbench", "-i", f"-s{scale}", connstr])
+        pg_bin.run_capture(["pgbench", "-i", "-I", "dtGvp", f"-s{scale}", connstr])
         pg_bin.run_capture(["pgbench", f"-T{n_restarts}", connstr])
 
     thread = threading.Thread(target=run_pgbench, args=(endpoint.connstr(),), daemon=True)

test_runner/regress/test_threshold_based_eviction.py

@@ -106,7 +106,7 @@ def test_threshold_based_eviction(
 
     # create a bunch of layers
     with env.endpoints.create_start("main", tenant_id=tenant_id) as pg:
-        pg_bin.run(["pgbench", "-i", "-s", "3", pg.connstr()])
+        pg_bin.run(["pgbench", "-i", "-I", "dtGvp", "-s", "3", pg.connstr()])
         last_flush_lsn_upload(env, pg, tenant_id, timeline_id)
     # wrap up and shutdown safekeepers so that no more layers will be created after the final checkpoint
     for sk in env.safekeepers: