MDEV-4506: Parallel replication: After-review fixes.

1 files changed, 69 insertions, 29 deletions

diff --git a/sql/rpl_parallel.cc b/sql/rpl_parallel.cc
index def0fe7c756..ce3170bb774 100644
--- a/sql/rpl_parallel.cc
+++ b/sql/rpl_parallel.cc
@@ -132,7 +132,6 @@ handle_rpl_parallel_thread(void *arg)
     while (!(events= rpt->event_queue) && !rpt->stop && !thd->killed)
       mysql_cond_wait(&rpt->COND_rpl_thread, &rpt->LOCK_rpl_thread);
     /* Mark that this thread is now executing */
-    rpt->free= false;
     rpt->event_queue= rpt->last_in_queue= NULL;
     thd->exit_cond(old_msg);
 
@@ -216,12 +215,24 @@ handle_rpl_parallel_thread(void *arg)
       {
         in_event_group= false;
 
+        /*
+          Remove any left-over registration to wait for a prior commit to
+          complete. Normally, such wait would already have been removed at
+          this point by wait_for_prior_commit(), but eg. in error case we
+          might have skipped waiting, so we would need to remove it explicitly.
+        */
         rgi->commit_orderer.unregister_wait_for_prior_commit();
         thd->wait_for_commit_ptr= NULL;
 
         /*
-          Record that we have finished, so other event groups will no
-          longer attempt to wait for us to commit.
+          Record that this event group has finished (eg. transaction is
+          committed, if transactional), so other event groups will no longer
+          attempt to wait for us to commit. Once we have increased
+          entry->last_committed_sub_id, no other threads will execute
+          register_wait_for_prior_commit() against us. Thus, by doing one
+          extra (usually redundant) wakeup_subsequent_commits() we can ensure
+          that no register_wait_for_prior_commit() can ever happen without a
+          subsequent wakeup_subsequent_commits() to wake it up.
 
           We can race here with the next transactions, but that is fine, as
           long as we check that we do not decrease last_committed_sub_id. If
@@ -246,6 +257,11 @@ handle_rpl_parallel_thread(void *arg)
     mysql_mutex_lock(&rpt->LOCK_rpl_thread);
     if ((events= rpt->event_queue) != NULL)
     {
+      /*
+        Take next group of events from the replication pool.
+        This is faster than having to wakeup the pool manager thread to give us
+        a new event.
+      */
       rpt->event_queue= rpt->last_in_queue= NULL;
       mysql_mutex_unlock(&rpt->LOCK_rpl_thread);
       goto more_events;
@@ -254,7 +270,7 @@ handle_rpl_parallel_thread(void *arg)
     if (!in_event_group)
     {
       rpt->current_entry= NULL;
-      if (!rpt->stop && !rpt->free)
+      if (!rpt->stop)
       {
         mysql_mutex_lock(&rpt->pool->LOCK_rpl_thread_pool);
         list= rpt->pool->free_list;
@@ -263,7 +279,6 @@ handle_rpl_parallel_thread(void *arg)
         if (!list)
           mysql_cond_broadcast(&rpt->pool->COND_rpl_thread_pool);
         mysql_mutex_unlock(&rpt->pool->LOCK_rpl_thread_pool);
-        rpt->free= true;
       }
     }
   }
@@ -300,6 +315,7 @@ rpl_parallel_change_thread_count(rpl_parallel_thread_pool *pool,
   uint32 i;
   rpl_parallel_thread **new_list= NULL;
   rpl_parallel_thread *new_free_list= NULL;
+  rpl_parallel_thread *rpt_array= NULL;
 
   /*
     Allocate the new list of threads up-front.
@@ -307,10 +323,13 @@ rpl_parallel_change_thread_count(rpl_parallel_thread_pool *pool,
     to allocate, and will not be left with a half-functional thread pool.
   */
   if (new_count &&
-      !(new_list= (rpl_parallel_thread **)my_malloc(new_count*sizeof(*new_list),
-                                                    MYF(MY_WME))))
+      !my_multi_malloc(MYF(MY_WME|MY_ZEROFILL),
+                       &new_list, new_count*sizeof(*new_list),
+                       &rpt_array, new_count*sizeof(*rpt_array),
+                       NULL))
   {
-    my_error(ER_OUTOFMEMORY, MYF(0), (int(new_count*sizeof(*new_list))));
+    my_error(ER_OUTOFMEMORY, MYF(0), (int(new_count*sizeof(*new_list) +
+                                          new_count*sizeof(*rpt_array))));
     goto err;;
   }
 
@@ -318,28 +337,16 @@ rpl_parallel_change_thread_count(rpl_parallel_thread_pool *pool,
   {
     pthread_t th;
 
-    if (!(new_list[i]= (rpl_parallel_thread *)my_malloc(sizeof(*(new_list[i])),
-                                                        MYF(MY_WME))))
-    {
-      my_error(ER_OUTOFMEMORY, MYF(0), sizeof(*(new_list[i])));
-      goto err;
-    }
+    new_list[i]= &rpt_array[i];
     new_list[i]->delay_start= true;
-    new_list[i]->running= false;
-    new_list[i]->stop= false;
-    new_list[i]->free= false;
     mysql_mutex_init(key_LOCK_rpl_thread, &new_list[i]->LOCK_rpl_thread,
                      MY_MUTEX_INIT_SLOW);
     mysql_cond_init(key_COND_rpl_thread, &new_list[i]->COND_rpl_thread, NULL);
     new_list[i]->pool= pool;
-    new_list[i]->current_entry= NULL;
-    new_list[i]->event_queue= NULL;
-    new_list[i]->last_in_queue= NULL;
     if (mysql_thread_create(key_rpl_parallel_thread, &th, NULL,
                             handle_rpl_parallel_thread, new_list[i]))
     {
       my_error(ER_OUT_OF_RESOURCES, MYF(0));
-      my_free(new_list[i]);
       goto err;
     }
     new_list[i]->next= new_free_list;
@@ -364,6 +371,13 @@ rpl_parallel_change_thread_count(rpl_parallel_thread_pool *pool,
     mysql_mutex_unlock(&LOCK_active_mi);
   }
 
+  /*
+    Grab each old thread in turn, and signal it to stop.
+
+    Note that since we require all replication threads to be stopped before
+    changing the parallel replication worker thread pool, all the threads will
+    be already idle and will terminate immediately.
+  */
   for (i= 0; i < pool->count; ++i)
   {
     rpl_parallel_thread *rpt= pool->get_thread(NULL);
@@ -381,7 +395,6 @@ rpl_parallel_change_thread_count(rpl_parallel_thread_pool *pool,
     mysql_mutex_unlock(&rpt->LOCK_rpl_thread);
     mysql_mutex_destroy(&rpt->LOCK_rpl_thread);
     mysql_cond_destroy(&rpt->COND_rpl_thread);
-    my_free(rpt);
   }
 
   my_free(pool->threads);
@@ -409,7 +422,6 @@ err:
   {
     while (new_free_list)
     {
-      rpl_parallel_thread *next= new_free_list->next;
       mysql_mutex_lock(&new_free_list->LOCK_rpl_thread);
       new_free_list->delay_start= false;
       new_free_list->stop= true;
@@ -421,8 +433,7 @@ err:
         mysql_cond_wait(&new_free_list->COND_rpl_thread,
                         &new_free_list->LOCK_rpl_thread);
       mysql_mutex_unlock(&new_free_list->LOCK_rpl_thread);
-      my_free(new_free_list);
-      new_free_list= next;
+      new_free_list= new_free_list->next;
     }
     my_free(new_list);
   }
@@ -471,6 +482,12 @@ rpl_parallel_thread_pool::destroy()
 }
 
 
+/*
+  Wait for a worker thread to become idle. When one does, grab the thread for
+  our use and return it.
+
+  Note that we return with the worker threads's LOCK_rpl_thread mutex locked.
+*/
 struct rpl_parallel_thread *
 rpl_parallel_thread_pool::get_thread(rpl_parallel_entry *entry)
 {
@@ -571,7 +588,7 @@ rpl_parallel::do_event(rpl_group_info *serial_rgi, Log_event *ev)
   rpl_parallel_entry *e;
   rpl_parallel_thread *cur_thread;
   rpl_parallel_thread::queued_event *qev;
-  rpl_group_info *rgi;
+  rpl_group_info *rgi= NULL;
   Relay_log_info *rli= serial_rgi->rli;
   enum Log_event_type typ;
 
@@ -596,6 +613,7 @@ rpl_parallel::do_event(rpl_group_info *serial_rgi, Log_event *ev)
         event_group_new_gtid(rgi, gtid_ev))
     {
       my_error(ER_OUT_OF_RESOURCES, MYF(MY_WME));
+      delete rgi;
       return true;
     }
     if ((rgi->deferred_events_collecting= rli->mi->rpl_filter->is_on()))
@@ -622,14 +640,33 @@ rpl_parallel::do_event(rpl_group_info *serial_rgi, Log_event *ev)
     }
     else
     {
-      /* Check if we already have a worker thread for this entry. */
+      /*
+        Check if we already have a worker thread for this entry.
+
+        We continue to queue more events up for the worker thread while it is
+        still executing the first ones, to be able to start executing a large
+        event group without having to wait for the end to be fetched from the
+        master. And we continue to queue up more events after the first group,
+        avoiding the overhead of worker threads constantly entering and
+        leaving the worker thread free list.
+
+        But if the worker thread is idle at any point, it may return to the
+        idle list or be servicing a different request. So check this, and
+        allocate a new thread if the old one is no longer processing for us.
+      */
       cur_thread= e->rpl_thread;
       if (cur_thread)
       {
         mysql_mutex_lock(&cur_thread->LOCK_rpl_thread);
         if (cur_thread->current_entry != e)
         {
-          /* Not ours anymore, we need to grab a new one. */
+          /*
+            The worker thread became idle, and returned to the free list and
+            possibly was allocated to a different request. This also means
+            that everything previously queued has already been executed, else
+            the worker thread would not have become idle. So we should
+            allocate a new worker thread.
+          */
           mysql_mutex_unlock(&cur_thread->LOCK_rpl_thread);
           e->rpl_thread= cur_thread= NULL;
         }
@@ -682,6 +719,9 @@ rpl_parallel::do_event(rpl_group_info *serial_rgi, Log_event *ev)
       Events like ROTATE and FORMAT_DESCRIPTION. Do not run in worker thread.
       Same for events not preceeded by GTID (we should not see those normally,
       but they might be from an old master).
+
+      The varuable `current' is NULL for the case where the master did not
+      have GTID, like a MariaDB 5.5 or MySQL master.
     */
     qev->rgi= serial_rgi;
     rpt_handle_event(qev, NULL);
@@ -719,8 +759,8 @@ rpl_parallel::do_event(rpl_group_info *serial_rgi, Log_event *ev)
   else
     cur_thread->event_queue= qev;
   cur_thread->last_in_queue= qev;
-  mysql_cond_signal(&cur_thread->COND_rpl_thread);
   mysql_mutex_unlock(&cur_thread->LOCK_rpl_thread);
+  mysql_cond_signal(&cur_thread->COND_rpl_thread);
 
   return false;
 }