New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 598379 link

Starred by 1 user

Issue metadata

Status: Verified
Owner:
Last visit > 30 days ago
Closed: Apr 2016
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 2
Type: Bug



Sign in to add a comment

dummy_PassServer.sanity : fail setting up container

Project Member Reported by gwendal@chromium.org, Mar 28 2016

Issue description

Some HW test suite fails very fast:
http://cautotest.corp.google.com/afe/#tab_id=view_job&object_id=58156823

In https://pantheon.corp.google.com/storage/browser/chromeos-autotest-results/58156823-chromeos-test/chromeos4-row4-rack9-host1/ssp_logs/debug/

We create the container through:
/usr/local/autotest/server/autoserv -p -r /usr/local/autotest/results/58156823-chromeos-test/chromeos4-row4-rack9-host1 -m chromeos4-row4-rack9-host1 -u chromeos-test -l expresso-release/R51-8121.0.0/sanity/dummy_PassServer.sanity -s --lab True -P 58156823-chromeos-test/chromeos4-row4-rack9-host1 -n /usr/local/autotest/results/drone_tmp/attach.17649 --require-ssp --parent_job_id 58156804 --verify_job_repo_url

But /usr/local/autotest/results/drone_tmp/attach.17649 does not exist [anymore?] when we try to copy it back for safekeeping. 
 

Comment 1 by dshi@chromium.org, Mar 28 2016

Cc: -dshi@chromium.org
Labels: -Pri-1 Pri-2
Owner: dshi@chromium.org
The drone_tmp folder is deleted when there is a scheduler restart. Ideally, the content of drone_tmp/attach.xxx should be picked up by the test once the test started, so the deletion of the attach file won't cause test failure. 

In the case of server side packaging, the problem is that attache file is not copied over to container until it's setup, which takes 20-30 seconds. So there is a window that scheduler restart can fail a test using server-side packaging.

Maybe we should read the file right after the test is started.

Meanwhile, it's a race condition and shouldn't happen a lot (we only restart scheduler during push to prod). 
Project Member

Comment 2 by bugdroid1@chromium.org, Mar 29 2016

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/third_party/autotest/+/38fff70fc5257cd1497455e7fbc2742a2bd47ba9

commit 38fff70fc5257cd1497455e7fbc2742a2bd47ba9
Author: Dan Shi <dshi@google.com>
Date: Mon Mar 28 20:53:18 2016

[autotest] Save control file before setting up container

This is to prevent test failure caused by missing control file.
Scheduler restart leads to all control files in drone_tmp being deleted.
Due to the long delay of setting up container when running tests with SSP,
this becomes more of an issue.

This change saves the control file to result folder before setting up
container.

BUG= chromium:598379 
TEST=local run test, unittest

Change-Id: I10a7715477ce14d3cbba5176292fc9aca19f9f2d
Reviewed-on: https://chromium-review.googlesource.com/335269
Commit-Ready: Dan Shi <dshi@chromium.org>
Tested-by: Dan Shi <dshi@chromium.org>
Reviewed-by: Gwendal Grignou <gwendal@chromium.org>

[modify] https://crrev.com/38fff70fc5257cd1497455e7fbc2742a2bd47ba9/site_utils/lxc.py

Comment 3 by dshi@chromium.org, Apr 17 2016

Status: Fixed (was: Untriaged)

Comment 4 by benhenry@google.com, Apr 27 2016

Components: Infra>Client>ChromeOS
Labels: -Infra-ChromeOS
Status: Verified (was: Fixed)
Bulk verified

Sign in to add a comment