On 20.06.2016 16:55, Sage Weil wrote:
On Mon, 20 Jun 2016, Igor Fedotov wrote:
Hi All,
I'm hitting the following issue while running Synthetic test case (from
store_test.cc) against bluestore:
store_test asserts on read data mismatch when read happens after the
zero-length write above object boundary.
Test case assumes all the data between the old and new object boundaries are
filled with zeros. But current bluestore write path implementation simply
discards zero-length writes and hence doesn't raise object size. As a result
subsequent read doesn't return anything above the old boundary.
The question is what's the proper fix for that:
1) Ignore zero-length writes in the test case and do not increase object
size/fill content on them.
or
2) Increase object content size on such writes in the bluestore.
Heh, I just hit this too. I think it makes more sense for 0
length writes to not affect object size. I added a test for it
and fixed Memstore. store_test.cc needs to be adjusted still though...
Pushed patches to top of wip-bluestore-2q, do you mind pulling those 2 out
and fixing store_test in another PR?
OK, will do.
Thanks!
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
