On 07/02/2024 3:10 pm, Anthony PERARD wrote: > On Wed, Feb 7, 2024 at 2:55 PM Andrew Cooper <[email protected]> > wrote: >> On 07/02/2024 2:34 pm, Anthony PERARD wrote: >>> On Wed, Feb 7, 2024 at 2:24 PM Andrew Cooper <[email protected]> >>> wrote: >>>>>> Stage: test >>>>>> Name: qemu-smoke-riscv64-gcc >>>>> I have to admit that I can't connect what was pushed recently to this job >>>>> failing. >>>> The qemu smoke tests for riscv and ppc intermittently fail on the >>>> OSSTest-lab infrastructure in Gitlab. >>>> >>>> We've never got to the bottom of it. >>> It's maybe because joubertin* and nocera* have more than one NUMA >>> nodes. It's look like 2 for nocera* and 4 for joubertin*. >>> Would that be enough of of an explanation about those tests taking a >>> bit longer than expected? >> That feels a bit like a blind guess. > Yes, but I think I did a test earlier, and the riscv test took more > like 10 seconds, but can also easily finish in less than a second. I > did the test a while ago, so I could misremember.
Ok - that is a wild variation. Something wonky is clearly going on. > >> What hardware are these? They're rather old, I seem to recall? > nocera: > http://logs.test-lab.xenproject.org/osstest/results/host/nocera1.examine/ AMD Opteron(tm) Processor 4133 a.k.a. Lisbon uarch from 2010. > joubertin: > http://logs.test-lab.xenproject.org/osstest/results/host/joubertin0.examine/ AMD Opteron(tm) Processor 6376 a.k.a. Abu-Dhabi uarch from 2012. > nocera have been here for a while, since before 2015. I don't really > know how to find out how old the machine is. joubertin* have never > been used with osstest so far, but the first mention of it I can find > of it is from 2017. They were ancient when added to OSSTest. ~Andrew
