See CASSANDRA_DESIGN.md. We added a shard_id to avoid
data loss on restarts. On restart the data of some
given datapoints may be spread on several cassandra cells.
Limit of fetch data queries were computed with no
restart, and then may be too short in case of restarts.
This may create artificial loss at read time.
Coverage increased (+0.008%) to 81.66% when pulling 46425a763db3f957f66202f77c39d2c4b4ec4872 on Thib17:bug_limit into b6ef4adfb09e9b9ce97fca409c563f26bac4cb99 on criteo:master.
See CASSANDRA_DESIGN.md. We added a shard_id to avoid data loss on restarts. On restart the data of some given datapoints may be spread on several cassandra cells.
Limit of fetch data queries were computed with no restart, and then may be too short in case of restarts. This may create artificial loss at read time.