[Bug 219399] System panics after several hours of 14-threads-compilation orgies using poudriere on AMD Ryzen...

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=219399

Mark Millard <markmi at dsl-only.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |markmi at dsl-only.net

--- Comment #13 from Mark Millard <markmi at dsl-only.net> ---
(In reply to Nils Beyer from comment #12)

As an odd data point: I use a Ryzen 1800X
system running FreeBSD head in VirtualBox
under Windows 10. I use it for buildworld
buildklernel for itself and for cross builds
to powerpc64, powerpc, aarch64, and armv6/7.
Also for building amd64 ports for itself.

It even happens to be a X370 Gaming Pro
Carbon (MS-7A32) context. But it has
lots of slower RAM instead of the more
typical less-but-faster RAM. It also
has very good processor cooling as I
understand.

I tend to use -j16 (so all the threads)
for buildworld buildkernel and can keep
the threads busy doing builds for all
the TARGET_ARCH's.

In this context I've not seen any such
problems other than when I first
started: The machine had been
"optimized" to the point of occasional
unreliability initially. Backing off
the memory handling configuration from
what it was when I was given access to
the machine fixed that issue. (Later
the initial RAM was replaced with even
more RAM.)

And, yes, FreeBSD buildworld buildkernel
seemed to push some issue harder than
other testing techniques had. (Not that
I have a detailed fault analysis.)

Note: I've not been using poudriere to
do any of the builds.

Not a great analogy to your context but
the above has some points in common
with your context.

-- 
You are receiving this mail because:
You are the assignee for the bug.