You mean like mb.py run? Because that's what is the closest to reproducing with a local build.

Otherwise you can fetch the binaries but it's likely less interesting for debugging.

I think the easiest is click on the link of the failing shard, then follow the instruction in the swarming dashboard:

"Run this task locally:
  python swarming.py reproduce -S https://chromium-swarm.appspot.com 316bf4469db56b10
"

I've seen the python swarming.py reproduce but that feels too end-to-end (as in: pulls all the binaries).
What I am missing here is pretty simple: what is the command line that this test executes?

For the other steps (e.g. compile, base_unittest) it is always visible in stdio.
What is the equivalent for telemetry_perf_unittests?
Very likely altimin@ (the author of the CL) wants to reproduce the failure locally against its build, so he can add printfs and more debugging things.
How should he proceed?

There is the command at the end of swarming output: 

"Command: /usr/bin/python ../../testing/scripts/run_telemetry_as_googletest.py ../../tools/perf/run_tests -v --xvfb --isolated-script-test-output=/b/swarm_slave/w/iosI6GaK/output.json"

It probably could be improve to make it clearer which command was run on the shard.

Though running the whole telemetry suite is probably idea since it takes very long. We have documentation for dealing with telemetry test failures in https://www.chromium.org/developers/telemetry/telemetry-unittests . Would be nice to add support linking to this doc in the https://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_rel_ng/builds/302496 page when the test fails.

agree with #5
would be great if each buildbot step had a "docs / help" link that pointed to some page like that.

This is less of swarming/buildbot request and more of the chromium recipe && individual test runners to emit better metadata in their steps, failing or otherwise.

This could be a dup of issue 715089?

I'm going to dup this with issue 715089.