MDEV-9162 : MariaDB Galera Cluster memory leak on async slave node

As galera node (slave) received query log events from an async
replication master, it partially wrote the updates made to replication
state table (mysql.gtid_slave_pos) to galera transaction writeset post
TOI. As a result, the transaction handle, thus created within galera,
was never freed/purged as the corresponding trx did not commit.
Thus, it kept piling up for every query log event and was only reclaimed
upon server shutdown when the transaction map object got destructed.
Fixed by making sure that updates in replication slave state table
are not written to galera transaction writeset and thus, not replicated
to other nodes.

4 files changed, 20 insertions, 3 deletions

diff --git a/sql/rpl_gtid.cc b/sql/rpl_gtid.cc
index 709cb284df5..81f2dd5812a 100644
--- a/sql/rpl_gtid.cc
+++ b/sql/rpl_gtid.cc
@@ -574,6 +574,14 @@ rpl_slave_state::record_gtid(THD *thd, const rpl_gtid *gtid, uint64 sub_id,
   if ((err= gtid_check_rpl_slave_state_table(table)))
     goto end;
 
+#ifdef WITH_WSREP
+  /*
+    Updates in slave state table should not be appended to galera transaction
+    writeset.
+  */
+  thd->wsrep_skip_append_keys= true;
+#endif
+
   if (!in_transaction)
   {
     DBUG_PRINT("info", ("resetting OPTION_BEGIN"));
@@ -687,6 +695,10 @@ IF_DBUG(dbug_break:, )
 
 end:
 
+#ifdef WITH_WSREP
+  thd->wsrep_skip_append_keys= false;
+#endif
+
   if (table_opened)
   {
     if (err || (err= ha_commit_trans(thd, FALSE)))
diff --git a/sql/sql_class.cc b/sql/sql_class.cc
index 5f1714558eb..6c7282c3505 100644
--- a/sql/sql_class.cc
+++ b/sql/sql_class.cc
@@ -962,6 +962,10 @@ extern "C" int wsrep_thd_retry_counter(THD *thd)
 {
   return(thd->wsrep_retry_counter);
 }
+extern "C" bool wsrep_thd_skip_append_keys(THD *thd)
+{
+  return thd->wsrep_skip_append_keys;
+}
 
 extern int
 wsrep_trx_order_before(void *thd1, void *thd2)
@@ -1083,7 +1087,8 @@ THD::THD()
    wsrep_po_cnt(0),
    wsrep_po_in_trans(FALSE),
    wsrep_apply_format(0),
-   wsrep_apply_toi(false)
+   wsrep_apply_toi(false),
+   wsrep_skip_append_keys(false)
 #endif
 {
   ulong tmp;
diff --git a/sql/sql_class.h b/sql/sql_class.h
index 596a2713d19..83899ecee21 100644
--- a/sql/sql_class.h
+++ b/sql/sql_class.h
@@ -3863,6 +3863,7 @@ public:
 #endif /*  GTID_SUPPORT */
   void*                     wsrep_apply_format;
   bool                      wsrep_apply_toi; /* applier processing in TOI */
+  bool                      wsrep_skip_append_keys;
 #endif /* WITH_WSREP */
 };
 
diff --git a/sql/wsrep_mysqld.h b/sql/wsrep_mysqld.h
index 4e0d2f4128f..1616bab695e 100644
--- a/sql/wsrep_mysqld.h
+++ b/sql/wsrep_mysqld.h
@@ -180,7 +180,7 @@ extern "C" query_id_t wsrep_thd_wsrep_last_query_id(THD *thd);
 extern "C" void wsrep_thd_set_wsrep_last_query_id(THD *thd, query_id_t id);
 extern "C" void wsrep_thd_awake(THD *thd, my_bool signal);
 extern "C" int wsrep_thd_retry_counter(THD *thd);
-
+extern "C" bool wsrep_thd_skip_append_keys(THD *thd);
 
 extern void wsrep_close_client_connections(my_bool wait_to_end);
 extern int  wsrep_wait_committing_connections_close(int wait_time);
@@ -195,7 +195,6 @@ extern bool wsrep_start_replication();
 extern bool wsrep_sync_wait (THD* thd, uint mask = WSREP_SYNC_WAIT_BEFORE_READ);
 extern int  wsrep_check_opts (int argc, char* const* argv);
 extern void wsrep_prepend_PATH (const char* path);
-/* some inline functions are defined in wsrep_mysqld_inl.h */
 
 /* Other global variables */
 extern wsrep_seqno_t wsrep_locked_seqno;