Closed holly-g closed 2 years ago
On going documentation for end-to-end generation of December reports with various checks
pip install -r requirements
make generate_parameters
make all -j 30
npm run build
python -m http.server
to push report data to the production bucket
make sync-prod
or to copy from the dev to prod
gsutil -m rsync -r -d gs://gtfs-data-test/report_gtfs_schedule/ gs://gtfs-data/report_gtfs_schedule/
If there are no changes between development and production rerun the last github action workflow run on main.
[x] verify in the github action that it is copying production report data
[x] verify that production site has been updated with most up to date month
obtain test emails from evan. Verify with Olivia and calitp that email contents are correct. Update the config file
[ ] update config file to have current month
[ ] verify from Olivia that email content is correct
script execution in development
python 3_generate_report_emails.py development
[ ] send out emails to evan's test.csv using sandbox token.
[ ] send out emails to test.csv using production token
[ ] verify that the emails pass visual inspection
Once emails have passed visual inspection, change config file for production
python 3_generate_report_emails.py production
pass prompt that asks if production is correct and verify email recipients are production send out emails
It's an odd flow that re-running the last github action on main
after pushing new reports data to the prod bucket is the way to get prod updated with the new report data. A minor improvement might be to augment the action to also support a workflow_dispatch
trigger with an env input
It's an odd flow that re-running the last github action on
main
after pushing new reports data to the prod bucket is the way to get prod updated with the new report data. A minor improvement might be to augment the action to also support aworkflow_dispatch
trigger with an env input
I've wondered if we even need GitHub actions for deployment. Maybe I'm living in the past, but a process where local development and deployment from the local machine occurs seems like it could be quicker than waiting ~1 hour to upload all the raw notebook data (ok maybe not an hour if you're only updating only 1 month) data and another ~10 minutes for a GitHub action to complete since it then needs to download all that raw data. Also, it's also multi-step deployment process that is prone to human error and distraction. If it's possible to upload just minified HTML/JS/CSS directly from a local machine, it seems like that could be much faster.
See https://github.com/cal-itp/reports/issues/66 as a place to continue the discussion.
December reports were sent out on 12-12-21