# 2020-09-23 Webapp Blank Building Page Postmortem

Date: 2020-09-23

Summary: The building page was inaccessible/blank for 1 hour & 2 minutes due to change introduced to support virtual tour for a building

Impact:

  • User's can't view a building as the building page was unable to render
  • Users can't proceed to book a viewing as the building page was inaccessible.

Root causes: In order to support virtual tour for a building, our webapp expects a key from algolia document. Due to that missing key, the building page got broken.

Resolution: Ran the refresh all records script to update the algolia document.

Detection: Manual + synthetic monitoring

Action items:

Action item Type
Fail build if there is any error while building static page prevent
Deploy the latest data first and don't break backward compatibility before making UI changes mitigate
Look for incident failure on priority process
Establish rotation plan in pager duty to fix flakiness process

# Lessons Learned

# What went well

  • Synthetic monitoring did fail & raised alert on our incident slack channel.
  • We spend minimum time in production to learn about the issue
  • Algolia refresh script is very fast and took very minimal time.

# What went wrong

  • No sanity on staging & prod after deployment
  • Acceptance tests raised no errors
  • Build did not fail while building pages although it had type error in the console.
  • Not actively looking into the incidents channel
  • Harsh going to production

# Where we got lucky

  • Not a lot of active users
  • Algolia refresh didn't required a rebuild of webapp. Don't know why ?

# Timeline

2020-09-23 (all times UTC)

  • 11:02 CircleCI deploys the latest change to production
  • 11:06 Synthetic monitoring raised an alert
  • 11:56 Harsh notifed on slack channel that prod is down
    • Harsh shared a screenshot in which building page was blank.
    • This is the error in the browser console: Cannot read property 'path' of undefined
    • Hard reloading didn't seem to help
  • 11:57 Aman replied on the thread by Harsh and investigated the error.
    • use of path was introduced for virtual tour, hence the following solutions came around. Either of them can be used to solve the problem:
      1. Revert the code change on the building page for virtual tour
      2. Run the algolia refresh
  • 12:01 Aman ran agolia refresh for staging & prod
  • 12:03 Aman did the sanity on staging & prod and verifed that everyhting is working.
  • 12:06 Incident & synthetic monitoring failure is resolved