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

feat(ha-term-safety): on_get_entry_term + on_truncate_from callback bodies (task 2.2)

Browse source 
on_get_entry_term → persist_entry_term (recorded term, 0=absent, legacy-0→1; caveat
documented per 1.3 review). on_truncate_from → persist_truncate_wal_above(p, index-1)
(delete LSN>=index), returns rc for library/leader retry; non-promotable fence intact;
replaces the committed-boundary heuristic on the library-driven path. New
truncate_term_index_locked keeps the in-memory per-LSN term index consistent on both
truncate paths (under lock, cutoff-aligned) → no stale-term read after truncation.
3 tests, reviewer APPROVED, per-group 0 failures.
This commit is contained in:
Rongsong Shen 2026-06-22 05:29:06 +00:00
parent 8f3d3ce8f4
commit ed2443533f
5 changed files with 210 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

55
src/controller/ha.c
View file

@ -1414,10 +1414,27 @@ static int ha_cb_persist_entry_stub(void *ctx, uint64_t index, uint64_t term,
return persist_wal_write_replicated(p, index, type, payload, plen);
}
/* on_get_entry_term (ha-term-safety task 2.2, R1.5/R3.1): return the recorded
* Raft term for the entry at 1-based LSN `index`, sourced from the WAL term
* index via persist_entry_term. Returns 0 when the LSN is absent (no entry).
* The library calls this for its logOk / (term,lsn) reconcile gate.
*
* Legacy-0 caveat (from the task 1.3 review MEDIUM): persist_entry_term maps a
* present legacy-0 entry (hdr.term == 0, written before this feature) to
* WAL_LEGACY_TERM (== 1). In a mixed pre/post-feature cluster where a
* post-feature leader is genuinely in term 1, a (term,lsn) EQUALITY check
* cannot tell a legacy-0 entry (mapped to 1) apart from a real term-1 entry by
* term alone. This is safe because the Raft log-matching invariant (same term
* + same index ⇒ same entry, enforced across the prefix) plus the bounded
* one-time upgrade window (every entry written after upgrade carries a real
* term) mean the only collision is at the legacy/term-1 boundary, where the
* prefixes already agree. We note it explicitly rather than special-casing it. */
static uint64_t ha_cb_get_entry_term_stub(void *ctx, uint64_t index)
{
(void)ctx; (void)index;
return 0; /* TODO task 2.2: 0 == absent (logOk reads local last-log term as 0) */
ha_state_t *ha = (ha_state_t *)ctx;
if (!ha || !ha->raft_c)
return 0; /* test mode without a cluster: no log → absent */
return persist_entry_term(&ha->raft_c->persist, index);
}
static int ha_cb_send_append_entries_stub(void *ctx, int peer_idx,
@ -1430,10 +1447,26 @@ static int ha_cb_send_append_entries_stub(void *ctx, int peer_idx,
return 0; /* TODO task 2.3 */
}
/* on_truncate_from (ha-term-safety task 2.2, R1.8/R3.2): delete the local log
* entries at LSN `index..end` because they conflict with the leader's log at a
* (term,lsn) the library detected. persist_truncate_wal_above(p, cutoff) removes
* every entry with LSN > cutoff, so to drop LSN >= index we pass cutoff =
* index - 1. This REPLACES the committed-boundary truncate heuristic on the
* library-driven path: the library, not a watermark, decides the cut point.
*
* Returns the truncate rc. A non-zero rc propagates so the library/leader
* retries; the member stays non-promotable per R3.5 (that fence is existing
* behavior and is not touched here). persist_truncate_wal_above also shrinks
* the per-LSN term index for the removed LSNs, so a later on_get_entry_term on
* a truncated LSN reads back 0 (absent). */
static int ha_cb_truncate_from_stub(void *ctx, uint64_t index)
{
(void)ctx; (void)index;
return 0; /* TODO task 2.4 */
ha_state_t *ha = (ha_state_t *)ctx;
if (!ha || !ha->raft_c)
return -EINVAL; /* test mode without a cluster: cannot truncate */
if (index == 0)
return -EINVAL; /* LSN 0 is reserved — never an entry to truncate */
return persist_truncate_wal_above(&ha->raft_c->persist, index - 1);
}
static void ha_cb_commit_advanced_stub(void *ctx, uint64_t commit_index)
@ -1633,6 +1666,20 @@ int ha_cb_persist_entry_for_test(ha_state_t *ha, uint64_t index, uint64_t term,
{
return s_real_cbs.on_persist_entry(ha, index, term, entry, len);
}
/* ha-term-safety task 2.2: drive the WIRED on_get_entry_term callback so a unit
* test exercises the real entry-term accessor. Exposed only under XCE_TESTING. */
uint64_t ha_cb_get_entry_term_for_test(ha_state_t *ha, uint64_t index)
{
return s_real_cbs.on_get_entry_term(ha, index);
}
/* ha-term-safety task 2.2: drive the WIRED on_truncate_from callback so a unit
* test exercises the real conflict truncation. Exposed only under XCE_TESTING. */
int ha_cb_truncate_from_for_test(ha_state_t *ha, uint64_t index)
{
return s_real_cbs.on_truncate_from(ha, index);
}
#endif
/* ha_check_heartbeat_nocluster() / ha_watchdog_thread() / ha_start_watchdog() /