lei.chen/myxce
1
0
Fork 1
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

test(ha-commit-engine): full regression green + legacy-path audit (task 6.1)

Browse source 
Fixed the 3 bug-411 baseline failures honestly: stale timeout
expectation updated to the reject-on-timeout contract (engine-wired
fixture, killing power retained); two healthy-primary gate fixtures
populated with standby_endpoints (gate logic untouched). Census: 46
groups / 467 tests / 0 failures standalone under default ASAN.
Audit (reviewer-confirmed, recorded in tasks.md): no production walAck
path reaches the deprecated acked_lsn>0 counting or n_active==0
short-circuit except the documented LSN-0 first entry. Leak inventory
for 6 groups documented as follow-up.
This commit is contained in:
Rongsong Shen 2026-06-12 03:32:19 +00:00
parent f0d6074d4e
commit 95a10cedf3
6 changed files with 785 additions and 295 deletions
Download patch file Download diff file Expand all files Collapse all files

45
src/controller/testsuite/test_p2_integration.c
View file

@ -267,23 +267,41 @@ static void test_single_node_committed_lsn_advances(void)
}
/* ------------------------------------------------------------------ */
/* Test 8: Req 4.2 — committed_lsn must NOT advance when sync times out.
/* Test 8: Req 4.2 (P2) / ha-commit-engine Req 2.1, 4.4 — committed_lsn
* must NOT advance when sync times out, and the write is REJECTED.
*
* Set up persist with two active standbys that never ACK.
* Use a very short sync_timeout_ms so the wait times out quickly.
* Verify: persist_wal_sync_append returns 0 (WAL local success),
* sync_rc == -ETIMEDOUT, and committed_lsn stays at 0.
* UPDATED by ha-commit-engine task 6.1: this test previously pinned the
* OLD accept-on-timeout contract (rc==0 with sync_rc==-ETIMEDOUT as a
* mere advisory). persist_wal_sync_append now rejects on timeout
* (returns -ETIMEDOUT) so no unconfirmed write reaches the caller as
* success (Req 4.4). The fixture is also upgraded to the engine-wired
* production path: cluster_init_phase2 always wires the commit engine,
* so the honest place to pin "committed_lsn unmoved on timeout" is the
* engine-owned path, not the deprecated ce==NULL legacy counting.
*
* Set up: 3-member cluster (self + two standbys that never ACK), engine
* wired exactly as cluster_init_phase2 does. A priming async append
* makes the sync entry land at LSN 1 (LSN 0 routes to the legacy branch
* — the accepted once-per-cluster-lifetime interim behaviour).
* Verify: rc == -ETIMEDOUT (write rejected), sync_rc == -ETIMEDOUT,
* and committed_lsn stays at 0 (engine never advanced it).
*/
static void test_committed_lsn_does_not_advance_on_timeout(void)
{
commit_engine_t ce;
persist_t* p = make_p2_persist("clsn_to");
TEST_ASSERT_NOT_NULL(p);
/* Add two standbys — they will never ACK, so wait times out */
/* Add two standbys — they will never ACK, so the wait times out */
static int ep_a, ep_b;
repl_add_standby(&p->repl, (void *)&ep_a, "sb_a", 7010);
repl_add_standby(&p->repl, (void *)&ep_b, "sb_b", 7011);
/* Wire the engine exactly as cluster_init_phase2 does for n_peers=3 */
TEST_ASSERT_EQUAL_INT(0, commit_engine_init(&ce, p, 3, 8));
repl_set_commit_engine(&p->repl, &ce);
/* Use a very short timeout so the test finishes quickly */
p->config.sync_timeout_ms = 40;
@ -293,19 +311,24 @@ static void test_committed_lsn_does_not_advance_on_timeout(void)
/* Install drop hook so repl_broadcast doesn't dereference dummy eps */
repl_broadcast_send_hook = p2_drop_send_hook;
/* Priming append: sync entry below lands at LSN 1 (engine-wired path) */
TEST_ASSERT_EQUAL_INT(0, persist_wal_async(p, WAL_JOB_UPDATE, "p", 1));
int sync_rc = 0;
int rc = persist_wal_sync_append(p, WAL_JOB_SUBMIT, "job", 3, &sync_rc);
repl_broadcast_send_hook = NULL;
/* WAL write must succeed locally */
TEST_ASSERT_EQUAL_INT(0, rc);
/* Sync must report timeout (standbys registered but never ACK) */
/* No per-entry majority within the deadline → the write is REJECTED
* (ha-commit-engine Req 4.4: timeout propagates to the caller) */
TEST_ASSERT_EQUAL_INT_MESSAGE(-ETIMEDOUT, rc,
"sync append without majority ACK must be rejected with -ETIMEDOUT");
TEST_ASSERT_EQUAL_INT(-ETIMEDOUT, sync_rc);
/* committed_lsn must not have advanced (Req 4.2) */
/* committed_lsn must not have advanced (Req 4.2; engine Req 2.1) */
uint64_t committed_after =
atomic_load_explicit(&p->committed_lsn, memory_order_acquire);
TEST_ASSERT_EQUAL_UINT64(0, committed_after);
TEST_ASSERT_EQUAL_UINT64_MESSAGE(0, committed_after,
"committed_lsn must stay unmoved when the entry never reaches majority");
commit_engine_fini(&ce);
teardown_p2_persist(p);
printf("test_committed_lsn_does_not_advance_on_timeout: PASS\n");
}