need less ports

.github/actions/run-python-test-set/action.yml

@@ -19,7 +19,7 @@ inputs:
   run_in_parallel:
     description: 'Whether to run tests in parallel'
     required: false
-    default: 'true'
+    default: 'false'
   save_perf_report:
     description: 'Whether to upload the performance report, if true PERF_TEST_RESULT_CONNSTR env variable should be set'
     required: false
@@ -171,7 +171,7 @@ runs:
           --junitxml=$TEST_OUTPUT/junit.xml \
           --alluredir=$TEST_OUTPUT/allure/results \
           --tb=short \
-          --verbose \
+          --verbose -k "test_forward or test_create_snapsh" -x \
           -rA $TEST_SELECTION $EXTRA_PARAMS
 
         if [[ "${{ inputs.save_perf_report }}" == "true" ]]; then

.github/workflows/build_and_test.yml

@@ -5,6 +5,7 @@ on:
     branches:
       - main
       - release
+      - tmp-repro
   pull_request:
 
 defaults:

libs/utils/src/lsn.rs

@@ -6,6 +6,7 @@ use std::ops::{Add, AddAssign};
 use std::path::Path;
 use std::str::FromStr;
 use std::sync::atomic::{AtomicU64, Ordering};
+use tracing::info;
 
 use crate::seqwait::MonotonicCounter;
 
@@ -239,6 +240,7 @@ impl MonotonicCounter<Lsn> for RecordLsn {
         let new_prev = self.last;
         self.last = lsn;
         self.prev = new_prev;
+        info!("advanced record lsn to {}/{}", self.last, self.prev);
     }
     fn cnt_value(&self) -> Lsn {
         self.last

pageserver/src/tenant/timeline.rs

@@ -1246,6 +1246,11 @@ impl Timeline {
 
                 state,
             };
+            info!(
+                "initialized lsrlsn to {}/{}",
+                disk_consistent_lsn,
+                metadata.prev_record_lsn().unwrap_or(Lsn(0))
+            );
             result.repartition_threshold = result.get_checkpoint_distance() / 10;
             result
                 .metrics

poetry.lock

@@ -1,4 +1,4 @@
-# This file is automatically @generated by Poetry 1.4.0 and should not be changed by hand.
+# This file is automatically @generated by Poetry 1.4.1 and should not be changed by hand.
 
 [[package]]
 name = "aiohttp"
@@ -1932,6 +1932,21 @@ pytest = [
     {version = ">=6.2.4", markers = "python_version >= \"3.10\""},
 ]
 
+[[package]]
+name = "pytest-repeat"
+version = "0.9.1"
+description = "pytest plugin for repeating tests"
+category = "main"
+optional = false
+python-versions = ">=2.7, !=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*, !=3.4.*"
+files = [
+    {file = "pytest-repeat-0.9.1.tar.gz", hash = "sha256:5cd3289745ab3156d43eb9c8e7f7d00a926f3ae5c9cf425bec649b2fe15bad5b"},
+    {file = "pytest_repeat-0.9.1-py2.py3-none-any.whl", hash = "sha256:4474a7d9e9137f6d8cc8ae297f8c4168d33c56dd740aa78cfffe562557e6b96e"},
+]
+
+[package.dependencies]
+pytest = ">=3.6"
+
 [[package]]
 name = "pytest-timeout"
 version = "2.1.0"
@@ -2597,4 +2612,4 @@ testing = ["func-timeout", "jaraco.itertools", "pytest (>=6)", "pytest-black (>=
 [metadata]
 lock-version = "2.0"
 python-versions = "^3.9"
-content-hash = "2515a9320c2960076012fbc036fb33c4f6a23515c8d143785931dc18c6722d91"
+content-hash = "ce00f0cf8735e850b1cfb29d8ddf8f66fab2395c589fe7e16becc774edcc7845"

pyproject.toml

@@ -34,6 +34,7 @@ types-psutil = "^5.9.5.4"
 types-toml = "^0.10.8"
 pytest-httpserver = "^1.0.6"
 aiohttp = "3.7.4"
+pytest-repeat = "^0.9.1"
 
 [tool.poetry.group.dev.dependencies]
 black = "^23.1.0"

test_runner/fixtures/neon_fixtures.py

@@ -77,7 +77,7 @@ DEFAULT_BRANCH_NAME: str = "main"
 DEFAULT_PG_VERSION_DEFAULT: str = "14"
 
 BASE_PORT: int = 15000
-WORKER_PORT_NUM: int = 1000
+WORKER_PORT_NUM: int = 10000
 
 
 def pytest_configure(config: Config):

test_runner/fixtures/utils.py

@@ -149,7 +149,7 @@ def get_scale_for_db(size_mb: int) -> int:
 
 
 ATTACHMENT_NAME_REGEX: re.Pattern = re.compile(  # type: ignore[type-arg]
-    r"regression\.diffs|.+\.(?:log|stderr|stdout|filediff|metrics|html)"
+    r"config|postgresql\.conf|regression\.diffs|.+\.(?:log|stderr|stdout|filediff|metrics|html)"
 )
 
 

test_runner/regress/test_compatibility.py

@@ -136,6 +136,7 @@ def test_backward_compatibility(
 
 @pytest.mark.xdist_group("compatibility")
 @pytest.mark.order(after="test_create_snapshot")
+@pytest.mark.repeat(1000)
 def test_forward_compatibility(
     test_output_dir: Path,
     port_distributor: PortDistributor,

test_runner/regress/test_wal_page_boundary_start.py

@@ -1,60 +0,0 @@
-import time
-from fixtures.neon_fixtures import NeonEnv, fork_at_current_lsn, wait_for_sk_commit_lsn_to_arrive_at_pageserver_last_record_lsn
-from fixtures.utils import query_scalar
-from fixtures.types import Lsn
-from fixtures.log_helper import log
-
-# Test compute start from WAL page boundary.
-def test_wal_page_boundary_start(neon_simple_env: NeonEnv):
-    env = neon_simple_env
-
-    env.neon_cli.create_branch("init")
-    pg = env.postgres.create_start("init")
-    tenant_id = pg.safe_psql("show neon.tenant_id")[0][0]
-    timeline_id = pg.safe_psql("show neon.timeline_id")[0][0]
-    with pg.cursor() as cur:
-        cur.execute("create table t(key int, value text)")
-        # measure how much space logical message takes
-        lsn_before = Lsn(query_scalar(cur, "select pg_current_wal_lsn()"))
-        log.info(f"current_lsn={lsn_before}")
-        # Non-transactional logical message doesn't write WAL, only XLogInsert's
-        # it, so use transactional. Which is a bit problematic as transactional
-        # necessitates commit record. Alternatively we can do smth like
-        #   select neon_xlogflush(pg_current_wal_insert_lsn());
-        # but isn't much better + that particular call complains on 'xlog flush
-        # request 0/282C018 is not satisfied' as pg_current_wal_insert_lsn skips
-        # page headers.
-        payload = "blahblah"
-        cur.execute(f"select pg_logical_emit_message(true, 'pref', '{payload}')")
-        lsn_after_by_curr_wal_lsn = Lsn(query_scalar(cur, "select pg_current_wal_lsn()"))
-        lsn_diff = lsn_after_by_curr_wal_lsn - lsn_before
-        logical_message_base = lsn_after_by_curr_wal_lsn - lsn_before - len(payload)
-        log.info(f"before {lsn_before}, after {lsn_after_by_curr_wal_lsn}, lsn diff is {lsn_diff}, base {logical_message_base}")
-
-        # and write logical message spanning exactly till the page boundary
-        lsn_before = Lsn(query_scalar(cur, "select pg_current_wal_lsn()"))
-        log.info(f"current_lsn={lsn_before}")
-        curr_lsn = Lsn(query_scalar(cur, "select pg_current_wal_lsn()"))
-        offs = int(curr_lsn) % 8192
-        till_page = 8192 - offs
-        payload_len = till_page - logical_message_base - 8 # not sure why 8 is here, it is deduced from experiments
-        log.info(f"current_lsn={curr_lsn}, offs {offs}, till_page {till_page}")
-
-        cur.execute(f"select pg_logical_emit_message(true, 'pref', 'f{'a' * payload_len}')")
-        supposedly_page_boundary = Lsn(query_scalar(cur, "select pg_current_wal_lsn()"))
-        log.info(f"supposedly_page_boundary={supposedly_page_boundary}")
-        assert int(supposedly_page_boundary) % 8192 == 0, "missed page boundary, bad luck"
-
-    pg.stop()
-
-    # wait for WAL arrival at sk is ensured by synchronous commit.
-    wait_for_sk_commit_lsn_to_arrive_at_pageserver_last_record_lsn(tenant_id, timeline_id, env.safekeepers, env.pageserver)
-    pg_new = env.postgres.create_start("init")
-    with pg_new.cursor() as cur:
-        cur.execute("select 42")
-        # We have had a bug with sending zero page header before first WAL
-        # record is written; give time for it to appear.
-        time.sleep(1)
-        # ensure that writing works
-        cur.execute("insert into t values (1, 'pear')")
-