[tarantool-patches] Re: [PATCH 3/3] Transaction support for applier

[Konstantin Osipov <kostja@tarantool.org>]

* Georgy Kirichenko <georgy@tarantool.org> [19/03/03 23:30]:
> @@ -271,14 +276,20 @@ txn_write_to_wal(struct txn *txn)
>  		return -1;
>  
>  	struct txn_stmt *stmt;
> -	struct xrow_header **row = req->rows;
> +	struct xrow_header **remote_row = req->rows;
> +	struct xrow_header **local_row = req->rows + txn->n_remote_rows;
>  	stailq_foreach_entry(stmt, &txn->stmts, next) {
>  		if (stmt->row == NULL)
>  			continue; /* A read (e.g. select) request */
> -		*row++ = stmt->row;
> +		if (stmt->row->replica_id != 0 &&
> +		    stmt->row->replica_id != instance_id)
> +			*remote_row++ = stmt->row;
> +		else
> +			*local_row++ = stmt->row;
>  		req->approx_len += xrow_approx_len(stmt->row);
>  	}
> -	assert(row == req->rows + req->n_rows);
> +	assert(remote_row == req->rows + txn->n_remote_rows);
> +	assert(local_row == req->rows + req->n_rows);

AFAIU this patch tests local rows by means of
replication_skip_conflict option. Could you please also test it
with local on_replace/before_replace triggers? Besides, now that
applier issues multi-statement transactions, we could add a test
for on_commit/on_rollback triggers fired from applier as well. We
could set such triggers from statement-level triggers.


-- 
Konstantin Osipov, Moscow, Russia, +7 903 626 22 32
http://tarantool.io - www.twitter.com/kostja_osipov