On Tue, May 15, 2018 at 12:02 PM, Florian Weimer <fweimer(a)redhat.com> wrote:
Lately, I've seen quite a few spurious build failures. Random
SIGBUS is
particularly common, and gcc reports that it cannot reproduce the SIGBUS in
a second compilation, which usually points to a kernel/hardware issue.
The latest problem was a hang during a build (on
buildvm-armv7-07.arm.fedoraproject.org), with this kernel:
Linux
buildvm-armv7-07.arm.fedoraproject.org 4.16.6-302.fc28.armv7hl+lpae #1
SMP Tue May 1 23:15:35 UTC 2018 armv7l armv7l armv7l GNU/Linux
This affects multiple builders, so I suspect a kernel issue, not dying
hardware because AFAIK, the machines are independent.
The issue also affects copying out the log files for Koji, so they probably
do not show the actual place of the hang.
There's a stability issue post upgrade, the upgrades moved the
underlying hypervisors to RHEL 7.5 and the build VMs to Fedora 28 at
the same time, the issue is known and is being investigated/worked
upon.
Peter