nautilus: os/bluestore: fix 'unused' calculation #34794
Conversation
|
@smithfarm can't merge this PR pls take a look |
smithfarm
added
nautilus-batch-1
|
@yuriw My bad, sorry. Should be mergeable now. |
src/test/objectstore/store_test.cc
Outdated
| TEST_P(StoreTestSpecificAUSize, ReproBug41901Test) { | ||
| if(string(GetParam()) != "bluestore") | ||
| return; | ||
| SetVal(g_conf(), "bluestore_debug_enforce_settings", "hdd"); |
There was a problem hiding this comment.
nautilus does not have the "bluestore_debug_enforce_settings" option, which is causing failures like http://qa-proxy.ceph.com/teuthology/yuriw-2020-04-30_22:32:53-rados-wip-yuri3-testing-2020-04-30-1557-nautilus-distro-basic-smithi/5003166/teuthology.log
src/test/objectstore/store_test.cc
Outdated
| if(string(GetParam()) != "bluestore") | ||
| return; | ||
| SetVal(g_conf(), "bluestore_debug_enforce_settings", "hdd"); | ||
| g_conf().apply_changes(nullptr); |
There was a problem hiding this comment.
@neha-ojha So, just drop the following lines from the commit?
SetVal(g_conf(), "bluestore_debug_enforce_settings", "hdd");
g_conf().apply_changes(nullptr);
There was a problem hiding this comment.
@smithfarm don't think that's enough, @ifed01 thoughts?
There was a problem hiding this comment.
@neha-ojha, @smithfarm - no, that's not enough - test case will fail when running from SSD. I can see two options here - remove the test case itself or backport commit(s) introducing "bluestore_debug_enforce_settings" setting. Personally I think removing test case in this backport is OK.
Fixes: https://tracker.ceph.com/issues/41901 Signed-off-by: Igor Fedotov <ifedotov@suse.com> (cherry picked from commit c91cc3a) Conflicts: src/test/objectstore/store_test.cc - omitted test case "TEST_P(StoreTestSpecificAUSize, ReproBug41901Test)" from the backport, because nautilus does not have the "bluestore_debug_enforce_settings" option
The processing logic which relies on 'unused' bitmap makes sense for bluestore setup where min alloc size is different from device block size. Now omitting if that's not true. Signed-off-by: Igor Fedotov <ifedotov@suse.com> (cherry picked from commit f960a01)
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn> Signed-off-by: Igor Fedotov <ifedotov@suse.com> (cherry picked from commit 9854972)
smithfarm
force-pushed
the
wip-45064-nautilus
branch
from
40a64a1
to
5d27594
Compare
|
@neha-ojha @ifed01 OK, I removed the test case completely and updated the commit message of the first commit to reflect that. |
backport tracker: https://tracker.ceph.com/issues/45064
backport of #33052
parent tracker: https://tracker.ceph.com/issues/41901
this backport was staged using ceph-backport.sh version 15.1.1.389
find the latest version at https://github.com/ceph/ceph/blob/master/src/script/ceph-backport.sh