Liren, should we do a quick retro on the recent DB bug in 0.29.18? I'm wondering if there was anything more I could have done to catch it before release

also okay if the answer is "no it was something we didn't know at that time" or "not worth since we are moving away from file based stuff"

Yes, we should do a quick retro. This is my fault. I think there are two reasons: 1. We did see the acceptance test failure. But I thought it was due to how the acceptance test was setup. Since the test could not be updated locally, I decided that it did not worth the time to fix it. I did not realize that it was a genuine production issue. 2. The ultimate solution was to deprecate the file-based migration systems. I was hoping that this would be done soon so that we don’t need to deal with any more issues related to the file-based system. That was kind of wishful thinking.

My fault too for not being more detailed too.. My takeaways are: • We should be better at respecting the build status. We've been lax with build failures. • We should always have automatic migration tests that migrate from an earlier version to the current version, even after we deprecate the file based migrations.