Looking at https://logs.chromium.org/v/?s=chromium%2Fbb%2Fchromium.linux%2FCast_Linux%2F48582%2F%2B%2Frecipes%2Fsteps%2Fcompile%2F0%2Fstdout ninja_wrapper.py is being run with -j 50

That implies a lot of trashing on a 16gb 8 core vm.

Also looking at the linker step looks like it needs more mem.

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
31976 chrome-+  20   0 23.848g 0.012t 8.285g S  20.9 80.3   1:05.74 ld.lld

Thanks for taking a look.

If I'm reading correctly, that's a Buildbot build referenced; what we're surprised by is why LUCI builds are so much slower than Buildbot builds, in particular the compile step, which is why we'd like a copy of the Buildbot bot to rule out machine differences. Is that feasible?

It does seem likely that linking is the difference between LUCI and Buildbot, but as to what's causing the difference is still unclear --- the LUCI machine doesn't seem to be facing memory pressure http://shortn/_NJTTcQfqmO (nor the Buildbot machine http://shortn/_3G8dPF3waQ), so we're pretty mystified.

I'll get one to you.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/chrome-golo/chrome-golo/+/ebf7a021ed2a55d989d525a9cceba5994f466362

commit ebf7a021ed2a55d989d525a9cceba5994f466362
Author: Peter Schmidt <pschmidt@google.com>
Date: Thu Mar 01 22:09:23 2018

vm176-m1 is yours.

Thanks very much! I'll update the bug when we're done testing so you can reclaim it.

Sorry, I forgot to mention I wanted this machine in chromium-swarm --- is it possible to convert vm176-m1?

Sure.  https://chromium-swarm.appspot.com/bot?id=vm176-m1&sort_stats=total%3Adesc

Thank you!

I'm done testing with this bot, please reclaim it whenever convenient, thanks!

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infradata/config/+/d3c7a10dc47f01a2bdbf5db6223671483652266c

commit d3c7a10dc47f01a2bdbf5db6223671483652266c
Author: Jao-ke Chin-Lee <jchinlee@google.com>
Date: Tue Mar 20 20:59:53 2018