App Fails and Retrospectives

06/05/2015 - 11:30 to 11:50
Stage 4

Session abstract: 

This is a series of the most interesting fuckup postmortems we did and lessons we learnt. In this session, we will talk about some stories on how we solved major refactoring issues in the past. We will also tell you how to prepare a refactoring.

Google Play backup

After weeks of testing, the QA team gave the GO to publish the new version on Google Play. It happened an evening before a long weekend. The staged rollout started at 10%. Except we missed something : test the upgrade from previous app version to the current one. If the app persists data, there may be some issues. Everyone was away except the product manager and a developer. The build script was not working correctly on the developer’s machine. It was a LONG night . But we learned a lot after this incident ...  join to learn how the weekend was saved. 

internationalization issues

The Zalando app title is "barcode scanner" in Russian.  Learn how a library's string resources can override your own strings .

The refactoring

We are switching to a new library to accomplish the same results but the developers say : “it is way cooler”. Let’s replace Guava Event Bus by Otto Event Bus. There was not much test coverage. A few weeks later, we noticed some features were broken. We made a few rules on measured improvements in speed, memory. For example Picasso had some memory issues. The image library transition from Picasso to Android-Universal-Image-Loader went smoothly. The app was published weeks after the planned date, stakeholders were not happy. We will talk how to plan and estimate refactorings .