Replies: 1 comment 1 reply
-
This sounds like a very dangerous feature to me. many of the redis commands (e.g. INCR) will be incorrectly executed on the replica and it'll not only have a partial view of the master's data, but also i can argue that the data is corrupted. |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
In #2045, i found an interesting idea.
Yes, it is useful for caching with mostly idempotent writes, otherwise, i often encounter a case,replicas may not successfully synchronize with master again when replicas disconnect and need full sync with mater and master always has heavy traffic, since client output buffer and replication backlog is limited, and
bgsave
takes much time. If users could accept dropping some data, enabling this feature may mitigate this bad case.Could we pick up this idea?
Beta Was this translation helpful? Give feedback.
All reactions