Google Workspace Directory Issues

Incident Report for WorkOS

Postmortem

Google Workspace Users Incorrectly Deleted Incident Report

Between 20:08 UTC on October 19, 2021 and 17:14 UTC on October 20, 2021, a subset of Google Workspace Directories were impacted by an issue that incorrectly removed users from their Directories and triggered corresponding deletion and update webhook events. We apologize for the negative impact this had on our customers.

Who was affected?

This issue impacted Google Workspace Directories with more than 100 users. All other Directory types were not impacted by this issue.

What happened?

On Monday, October 18th, 2021 at 16:55 UTC we deployed a change to Directory Sync that introduced a bug in our job re-queuing logic. For Google Workspace directories with multiple pages of users (where a page is 100 users), syncing the directory would enqueue a new job for each page of users. When these jobs ran the next day, the presence of multiple, simultaneously-executing jobs created a race condition which caused some users in the Directory to be incorrectly deleted inside WorkOS.

We released a fix at 17:14 UTC on October 20th, 2021 that prevented these forked synchronization jobs from being scheduled and re-synchronized all Google Workspace directories to restore the integrity of our data.

What will we do to mitigate problems like this in the future?

We are taking steps to further ensure that multiple synchronization jobs cannot exist for a single Directory to prevent this race condition from reoccurring.

Investigation into this issue was complicated by updated ETag properties on users from Google registering as a change to the user’s data. This caused extraneous dsync.user.updated events to be issued by us. We have updated our change reconciliation to ignore this ETag.

Posted Oct 28, 2021 - 13:04 EDT

Resolved

All Google Workspace Directories should now be functioning as expected.
Posted Oct 20, 2021 - 19:17 EDT

Monitoring

We're continuing to monitor impacted Directories.
Posted Oct 20, 2021 - 17:54 EDT

Identified

We've identified why the issue has happened, and we're in the process of fixing it.
Posted Oct 20, 2021 - 16:18 EDT

Investigating

We've received reports of Google Workspace Directories receiving erroneous user suspended events. Our team is currently investigating the cause.
Posted Oct 20, 2021 - 13:12 EDT