[Gluster-devel] namespace.t fails with brick multiplexing enabled
amukherj at redhat.com
Wed Mar 7 11:57:52 UTC 2018
On Wed, Mar 7, 2018 at 7:38 AM, Varsha Rao <varao at redhat.com> wrote:
> Hello Atin,
> On Tue, Mar 6, 2018 at 10:23 PM, Atin Mukherjee <amukherj at redhat.com>
> > Looks like the failure is back again. Refer
> > https://build.gluster.org/job/regression-test-with-multiplex/663/console
> > this has been failing in other occurrences too.
> Is the core dump file available for this? As on testing locally, the
> namespace test does not fail.
The test didn't generate any core. It failed functionally.
*09:32:12* not ok 15 Got "N" instead of "Y", LINENUM:85*09:32:12*
FAILED COMMAND: Y check_samples CREATE 28153613 /namespace/foo patchy3
You may have to file a bug under gluster-infrastructure component asking
for a machine to debug this further given this is locally not reproducible.
> > On Mon, Feb 26, 2018 at 2:58 PM, Varsha Rao <varao at redhat.com> wrote:
> >> Hi Atin,
> >> On Mon, Feb 26, 2018 at 12:18 PM, Atin Mukherjee <amukherj at redhat.com>
> >> wrote:
> >> > Hi Varsha,
> >> >
> >> > Thanks for your first feature "namespace" in GlusterFS! As we run a
> >> > periodic
> >> > regression jobs with brick multiplexing, we have seen that
> >> > tests/basic/namespace.t fails constantly with brick multiplexing
> >> > enabled. I
> >> > just went through the function check_samples () in the test file and
> >> > looked to me the function was written with an assumption that every
> >> > process
> >> > will be associated with one brick instance and will have one log file
> >> > which
> >> > is not the case for brick multiplexing  . If you need further
> >> > question on
> >> > brick multiplexing, feel free to ask.
> >> Yes, it was written with that assumption. I will send a patch to fix it.
> >> Thanks,
> >> Varsha
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Gluster-devel