Nack. Actually they should -- a backup is intended to be a drop in replacement. So it needs to be up-to-date.

If we don't trust not-in-use backup servers enough to be just able to put them in use without actual provision / manual checks, then I suggest that we drop the 'backup' state completely.

Agree with prathmesh

I believe my suggestion is just a return to prior behavior. I noticed the problem because the push failed on chromeos-server8 due to "unstable services" scheduler / host_scheduler (e.g. they were crashlooping and deploy_server_local detected this). Crashlooping scheduler / host_scheduler has been the WAI behavior for a long time.

I'm fine with dropping backup servers from existence too, which accomplishes the same thing as this bug wants.

Do we want to keep a backup scheduler?

As I said, I'm also happy with backup roles being eliminated entirely.

Let's not bikeshed this though. Either do that, or stop pushing to backups. I have a CL for the latter here: https://chromium-review.googlesource.com/c/chromiumos/third_party/autotest/+/747464

In that case, I'm +1 removing the backup scheduler, instead of keeping it there but letting it get out of date.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/third_party/autotest/+/9ec24b5ae4c5c917b0166a4ebc6cf359b4a171b6

commit 9ec24b5ae4c5c917b0166a4ebc6cf359b4a171b6
Author: Aviv Keshet <akeshet@chromium.org>
Date: Wed Nov 01 00:31:09 2017

autotest: deploy_server skip backup servers

BUG= chromium:780105 
TEST=None

Change-Id: Ice35e829dbb8ad5496b239f5f0745ebd6a3bb294
Reviewed-on: https://chromium-review.googlesource.com/747464
Commit-Ready: Aviv Keshet <akeshet@chromium.org>
Tested-by: Aviv Keshet <akeshet@chromium.org>
Reviewed-by: Shuqian Zhao <shuqianz@chromium.org>

[modify] https://crrev.com/9ec24b5ae4c5c917b0166a4ebc6cf359b4a171b6/site_utils/deploy_server.py