The problem/use-case that the feature addresses
In one of redis documents(https://redis.io/commands/lastsave/), it recommends to record the timestamp returned by LASTSAVE before a BGSAVE command, and check the timestamp given by the LASTSAVE command regularly afterwards, if they're different, then prior backup job is done.
However, Linux systems do time synchronization, when a system's time is synchronized by the time synchronization service(e.g., systemd-timesyncd), timestamps before and after the BGSAVE command could be the same. In such a situation the below recommendation does not work.
issuing a BGSAVE command and checking at regular intervals every N seconds if LASTSAVE changed
Description of the feature
- Maybe Redis cloud have a
LASTSAVEUNIXso that we have a timestamp in milliseconds instead of seconds, so that when time is synchronized, we have little chanceLASTSAVEUNIXgives identical results.