MDEV-37662: Binlog Corruption When tmpdir is Full

The binary log could be corrupted when committing a large transaction
(i.e. one whose data exceeds the binlog_cache_size limit and spills
into a tmp file) in binlog_format=row if the server's --tmp-dir is
full. The corruption that happens is only the GTID of the errored
transaction would be written into the binary log, without any
body/finalizing events.  This would happen because the content of the
transaction wasn't flushed at the proper time, and the transaction's
binlog cache data was not durable while trying to copy the content
from the binlog cache file into the binary log itself. While switching
the tmp file from a WRITE_CACHE to a READ_CACHE, the server would see
there is still data to flush in the cache, and first try to flush it.
This is not a valid time to flush that data to the temporary file
though, as the GTID event has already been written directly to the
binary log.  So if this flushing fails, it leaves the binary log in a
corrupted state.

The flush itself is expected to happen in
THD::binlog_flush_pending_rows_event(). However, if there is no pending
event, the flush is skipped.

This patch fixes this issue by still flushing the tmp file to disk in
THD::binlog_flush_pending_rows_event() when there is no pending event.

Reviewed-by: TODO

Author: bnestere

mysql-test/suite/rpl/r/rpl_row_binlog_tmp_file_flush_enospc.result

@@ -0,0 +1,38 @@
+include/master-slave.inc
+[connection master]
+connection master;
+set @old_binlog_cache_size= @@global.binlog_cache_size;
+set @@global.binlog_cache_size=4096;
+#
+# Initialize test data
+connection master;
+create table t1 (a int, b longtext default NULL) engine=innodb;
+#
+# Create transaction with cache data larger than the binlog_cache_size
+# so it spills into a tmp file, then simulate ENOSPC while flushing
+# the tmp file.
+#
+set @@session.debug_dbug="+d,simulate_binlog_tmp_file_no_space_left_on_flush";
+insert into t1 values (2, repeat("y", 8192));
+ERROR HY000: Error writing file '/home/brandon/workspace/server_1011/build/mysql-test/var/tmp/mysqld.1/#sql/fd=83' (Errcode: 28 "No space left on device")
+set @@session.debug_dbug="";
+#
+# Create another transaction to make sure the server/replication can
+# continue working normally after the error
+#
+insert into t1 values (3, repeat("z", 8192));
+include/save_master_gtid.inc
+connection slave;
+include/sync_with_master_gtid.inc
+include/diff_tables.inc [master:test.t1,slave:test.t1]
+#
+# Cleanup
+connection master;
+drop table t1;
+include/save_master_gtid.inc
+connection slave;
+include/sync_with_master_gtid.inc
+connection master;
+set @@global.binlog_cache_size= @old_binlog_cache_size;
+include/rpl_end.inc
+# End of rpl_row_binlog_tmp_file_flush_enospc.test

sql/log.cc

@@ -6566,6 +6566,23 @@ int THD::binlog_flush_pending_rows_event(bool stmt_end, bool is_transactional)
     error= mysql_bin_log.flush_and_set_pending_rows_event(this, 0,
                                                           is_transactional);
   }
+  else
+  {
+    /*
+      There's no pending event, but we still need to flush the cache
+    */
+    binlog_cache_mngr *cache_mngr=
+        (binlog_cache_mngr *) thd_get_ha_data(this, binlog_hton);
+    if (cache_mngr)
+    {
+      DBUG_EXECUTE_IF("simulate_binlog_tmp_file_no_space_left_on_flush",
+                      { DBUG_SET("+d,simulate_file_write_error"); });
+      error=
+          flush_io_cache(cache_mngr->get_binlog_cache_log(is_transactional));
+      DBUG_EXECUTE_IF("simulate_binlog_tmp_file_no_space_left_on_flush",
+                      { DBUG_SET("-d,simulate_file_write_error"); });
+    }
+  }
 
   DBUG_RETURN(error);
 }