On Fri, 2 Sep 2016, Ramesh Chander wrote:
> I am trying to run ceph_test_objectstore with my changes and I see this
> intermittent failure.
>
>
> Store_test.cc At line 1187:
>
> 1174 ObjectStore::Transaction t;
> 1175 bufferlist bl;
> 1176 bl.append("abcde");
> 1177 t.write(cid, hoid, 0, 5, bl);
> 1178 cerr << "Append 5 bytes" << std::endl;
> 1179 r = apply_transaction(store, &osr, std::move(t));
> 1180 ASSERT_EQ(r, 0);
> 1181
> 1182 struct store_statfs_t statfs;
> 1183 int r = store->statfs(&statfs);
> 1184 ASSERT_EQ(r, 0);
> 1185 ASSERT_EQ(5, statfs.stored);
> 1186 ASSERT_EQ(0x10000, statfs.allocated);
> 1187 ASSERT_EQ(available0 - 0x10000, statfs.available);
>
> The total allocated are 1MB more than expected by this test.
>
> (gdb) p statfs
> $50 = {total = 10737418240, available = 10722672640, allocated = 65536, stored = 5, compressed = 0, compressed_allocated = 0, compressed_original = 0}
> (gdb) p available0
> $51 = 10723786752
> (gdb) p 10723786752-10722672640
> $52 = 1114112
> (gdb) p 1114112-65536
> $53 = 1048576
> (gdb) p $53 / 1024
> $54 = 1024
> (gdb) list
>
> I traced call and looks like there is really a 1MB allocation from
> BlueFS: _flush_range function.
>
> Is it possible that when we write an object BlueFS also allocated
> another 1MB block and causing this difference?
>
> How to avoid this failure? Is it test case issue ?
We merged a PR a day or two ago that disabled the allocated checks in the
unit test. If you rebase on master this should go away!
sage
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
