Clone this repo:
  1. c27ea9a Adding octopus board to platform test suites. by Sridhar Sonti · 14 hours ago master
  2. c24dd02 Merge "suite-scheduler: Support qs_pool and qs_account in scheduling." by Xixuan Wu · 2 days ago
  3. 028f673 suite-scheduler: Support qs_pool and qs_account in scheduling. by Xixuan Wu · 2 days ago
  4. bad9401 Merge "suite-scheduler: Add mapping from (build, suite, model) to override settings." by Xixuan Wu · 2 days ago
  5. 48e70d8 [suite_scheduler] Adding a new suite for EnrollRetainment. by Ruben Zakarian · 4 days ago


Due to, this must be run outside the chroot.

  1. Install Google Cloud SDK:

    1. Follow the instructions.

    2. Install the Python App Engine component:

       sudo apt-get install google-cloud-sdk-app-engine-python
  2. Log in to gcloud: gcloud auth login

  3. Install pip:

     sudo apt-get install python-pip
  4. Run python to download third-party libraries and credentials.

Known Issues

  1. Failed to install third-packages in requirement.txt:

    Workround: Run pip install -t lib -r requirements.txt locally on your workstation to verify whether it could work.

  2. Failed to download credentials from ‘gs://’:

    Workround: Run gsutil ls gs:// locally on your workstation to verify whether you can fetch credentials with gsutil.

    If there's any auth issue, it may be because

    a. You‘re not using your corp account to login gcloud. Please run gcloud auth list to check what’s the used account for gcloud.

    b. Your corp account is not whitelisted by suite-scheduler project. Please contact @xixuan to whitelist your ldap.

How to test:

  1. Presubmit will kick off sanity test for verifying suite_scheduler.ini:

     python --test_type sanity
  2. Run unittest:

     python' or 'python --test_type unittest
  3. Run integration test:

     python --test_type integration
  4. Run tests with debug mode:

     python --debug
  5. Run suite scheduler v2 as a local development server:

    If you want to verify any APIs locally, you can run suite scheduler v2 as a local development server (

    Assuming you already got credentials/cloud_sql_credentials.txt by running ‘’, set up some variables based on the database instance you want to connect to:

       # Set up CONNECTION_NAME based on the cloudsql project/instance
       # name.  For more details, see:
       # Set up USER based on the user you want to log in.  If you use
       # readonly_nopw, grab the password from (as of April, 2018):
       USER='readonly_nopw' # [CIDB username]

    a. Test locally on your workstation to make sure you're whitelisted:

       /google/data/ro/projects/speckle/mysql \
           -S /cloudsql/${CONNECTION_NAME} \
           -u ${USER} \
           || echo "You may need to run prodaccess"

    b. Start the cloud_sql_proxy in a new window to let your service talk to cloud sql. Note that cloud_sql_proxy needs to be downloaded from Google's public cloud website:

       # Create the /cloudsql directory, since the suite scheduler assumes
       # that this is where to find it's socket.
       sudo mkdir /cloudsql
       sudo chmod 1777 /cloudsql
       # Start the cloud sql proxy.
       ./cloud_sql_proxy \
           -dir=/cloudsql \
           -instances=${CONNECTION_NAME} \
           -credential_file=./credentials/suite-scheduler_client_secret_service_account.json &

    c. Finally, start in a window: app.yaml \
           --port=8888 \
           --admin_port=8001 \

    d. Test it:

       curl 'http://localhost:8888/cron/trigger_event', or open it in chrome.

    Please note that once a dev_appserver is started, it simulates data_store & task_queue, which means:

    a. The datastore doesn‘t contain any previous cron run’s information, so the first round of cron/trigger_event won't trigger any event.

    b. The task queue is newly created with 0 tasks in it unless cron/trigger_event adds tasks in it. Dev_appserver restart will cause the local task queue to get purged.