Is it intentional to read using a 10 second exact staleness? While this could lead to lower read latency, it will by definition read 10 seconds old data. The default would be to use a strong read which is guaranteed to return all changes that have been committed, which is also one of the USP's of Spanner compared to storage systems that use eventual consistency. Another option here could also be to use a max staleness of 10 seconds, instead of exact staleness, and let Spanner choose the optimal read timestamp.
Originally posted by @olavloite in #3 (comment)
Is it intentional to read using a 10 second exact staleness? While this could lead to lower read latency, it will by definition read 10 seconds old data. The default would be to use a strong read which is guaranteed to return all changes that have been committed, which is also one of the USP's of Spanner compared to storage systems that use eventual consistency. Another option here could also be to use a max staleness of 10 seconds, instead of exact staleness, and let Spanner choose the optimal read timestamp.
Originally posted by @olavloite in #3 (comment)