From: Piotr Dałek <piotr.dalek@ts.fujitsu.com>
Date: Fri, 29 Jan 2016 10:38:53 +0000 (+0100)
Subject: tools/rados: fix wrong op/object sizes in rand/seq bench
X-Git-Tag: v10.0.4~111^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=refs%2Fpull%2F7432%2Fhead;p=ceph.git

tools/rados: fix wrong op/object sizes in rand/seq bench

Metadata contain correct information about last bench write run, yet
these are ignored in subsequent bench rand/seq runs, resulting in
wrong perf numbers and failures when not using --no-verify.

Signed-off-by: Piotr Dałek <piotr.dalek@ts.fujitsu.com>
---

diff --git a/src/common/obj_bencher.cc b/src/common/obj_bencher.cc
index a77febc6cb81..8e6169a98b92 100644
--- a/src/common/obj_bencher.cc
+++ b/src/common/obj_bencher.cc
@@ -230,6 +230,8 @@ int ObjBencher::aio_bench(
         cerr << "Must write data before running a read benchmark!" << std::endl;
       return r;
     }
+    object_size = prev_object_size;   
+    op_size = prev_op_size;           
   }
 
   char* contentsChars = new char[op_size];