[Gluster-devel] distinguish heal vs I/O

[Emmanuel Dreyfus]

In tests/basic/afr/self-heald.t test 67, we check that I/O is not
accounted as self-healing.  This is done by running a few dd in the
backround and calling glfsheal to check that files used by the dd
commands are not listed.

glfsheal will decide a file is "Possibly undergoing heal" if locking
fail with EAGAIN, which happens during heal, but also during I/O. As I
understand there is no way to distinguish the two, and the test
considers a file listed "Possibly undergoing heal" means a failure.

On Linux it seems to still pass, but on NetBSD the test cause suprious
failures. I therefore propose to consider "Possibly undergoing heal" is
not a failure, in that change:
http://review.gluster.org/9074

I would appreciate if someone could explain  me how the test pass on
Linux. And if the "Possibly undergoing heal" change should be restricted
to NetBSD, I can resubmit that.

-- 
Emmanuel Dreyfus
http://hcpnet.free.fr/pubz
manu at netbsd.org