diff --git a/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/BigdataRDFContext.java b/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/BigdataRDFContext.java
index d76aff0a9d..62c895cab7 100644
--- a/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/BigdataRDFContext.java
+++ b/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/BigdataRDFContext.java
@@ -1566,37 +1566,46 @@ private Void innerCall() throws Exception {
                     doQuery(cxn, new NullOutputStream());
 //                        success = true;
                 } else {
-                    doQuery(cxn, os);
+                    try {
+                        doQuery(cxn, os);
 //                        success = true;
-                  /*
-                   * GROUP_COMMIT: For mutation requests, calling flush() on the
-                   * output stream unblocks the client and allows it to proceed
-                   * BEFORE the write set of a mutation has been melded into a
-                   * group commit. This is only a problem for UPDATE requests.
-                   * 
-                   * The correct way to handle this is to allow the servlet
-                   * container to close the output stream. That way the close
-                   * occurs only after the group commit and when the control has
-                   * been returned to the servlet container layer.
-                   * 
-                   * There are some REST API methods (DELETE-WITH-QUERY,
-                   * UPDATE-WITH-QUERY) that reenter the API using a
-                   * PipedInputStream / PipedOutputStream to run a query (against
-                   * the last commit time) and pipe the results into a parser that
-                   * then executes a mutation without requiring the results to be
-                   * fully buffered. In order for those operations to not deadlock
-                   * we MUST flush() and close() the PipedOutputStream here (at
-                   * last for now - it looks like we probably need to execute those
-                   * REST API methods differently in order to support group commit
-                   * since reading from the lastCommitTime does NOT provide the
-                   * proper visibility guarantees when there could already be
-                   * multiple write sets buffered for the necessary indices by
-                   * other mutation tasks within the current commit group.)
-                   */
-                     if (os instanceof PipedOutputStream) {
-                     os.flush();
-                     os.close();
-                  }
+                    }
+                    catch(Exception e){
+                        log.error(e);
+                        throw e;
+                    }
+                    finally {
+                        /*
+                         * GROUP_COMMIT: For mutation requests, calling flush() on the
+                         * output stream unblocks the client and allows it to proceed
+                         * BEFORE the write set of a mutation has been melded into a
+                         * group commit. This is only a problem for UPDATE requests.
+                         *
+                         * The correct way to handle this is to allow the servlet
+                         * container to close the output stream. That way the close
+                         * occurs only after the group commit and when the control has
+                         * been returned to the servlet container layer.
+                         *
+                         * There are some REST API methods (DELETE-WITH-QUERY,
+                         * UPDATE-WITH-QUERY) that reenter the API using a
+                         * PipedInputStream / PipedOutputStream to run a query (against
+                         * the last commit time) and pipe the results into a parser that
+                         * then executes a mutation without requiring the results to be
+                         * fully buffered. In order for those operations to not deadlock
+                         * we MUST flush() and close() the PipedOutputStream here (at
+                         * last for now - it looks like we probably need to execute those
+                         * REST API methods differently in order to support group commit
+                         * since reading from the lastCommitTime does NOT provide the
+                         * proper visibility guarantees when there could already be
+                         * multiple write sets buffered for the necessary indices by
+                         * other mutation tasks within the current commit group.)
+                         */
+
+                        if (os instanceof PipedOutputStream) {
+                            os.flush();
+                            os.close();
+                        }
+                    }
                 }
                 if (log.isTraceEnabled())
                     log.trace("Query done.");
diff --git a/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/DeleteServlet.java b/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/DeleteServlet.java
index 960ba4d1a6..4dcf6749a7 100644
--- a/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/DeleteServlet.java
+++ b/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/DeleteServlet.java
@@ -338,12 +338,12 @@ public Void call() throws Exception {
                   final FutureTask<Void> ft = new FutureTask<Void>(
                         queryTask);
 
-                  // Submit query for evaluation.
-                  context.queryService.execute(ft);
-
                   // Reads on the statements produced by the query.
                   final InputStream is = newPipedInputStream(os);
 
+                  // Submit query for evaluation.
+                  context.queryService.execute(ft);
+
                   // Run parser : visited statements will be deleted.
                   rdfParser.parse(is, baseURI);
 
diff --git a/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/UpdateServlet.java b/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/UpdateServlet.java
index 927a8cde34..9f01fe050b 100644
--- a/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/UpdateServlet.java
+++ b/bigdata-core/bigdata-sails/src/java/com/bigdata/rdf/sail/webapp/UpdateServlet.java
@@ -441,12 +441,12 @@ public Void call() throws Exception {
 							final FutureTask<Void> ft = new FutureTask<Void>(
 									queryTask);
 
-							// Submit query for evaluation.
-							context.queryService.execute(ft);
-
 							// Reads on the statements produced by the query.
 							final InputStream is = newPipedInputStream(os);
 
+							// Submit query for evaluation.
+							context.queryService.execute(ft);
+
 							// Run parser : visited statements will be deleted.
 							rdfParser.parse(is, baseURI);