We are designing a module to subscribe and process data from Pub/Sub topics. In case of an application failure, we want to safely restart and resume processing from a specific point in time. We're considering using seek based on timestamp to retrieve messages from the topic. However, the documentation mentions potential issues with clock skew (https://cloud.google.com/pubsub/docs/replay-overview#seek_to_a_time). How significant is this impact likely to be?
If we seek to a timestamp with a small positive or negative delta (e.g., a few minutes before or after the specified time), will we still retrieve all relevant data, or is there a risk of missing messages?
For example, if event E1 has a publishTime of T5, and we seek to T3 or T4 (a few minutes before), can we expect to receive E1? Is it safe to rely on seek based on timestamp for recovery?
LinkedIn
Twitter
