Why Redis doesn't delete rdb files when snapshot is disabled?

[singku]

rdb files are created if rdb persistence is enabled.
Redis will load rdb files during start up.
If Redis has rdb persistence enabled and created rdb file. But later persistence is disabled. the rdb file exist on disk. What if redis restarted later? Would redis load outdated database from that rdb file? If I disabled persistence, I'd expect redis to restart with empty keys. Is it more reasonable to let redis delete the rdb file once persistence is disabled?

[sundb]

1. If you turn off RDB and turn on aof before restarting, data may be lost, please see [BUG] Complete data loss when enabling AOF through config #12484
2. If you want to restart with empty keys, please backup the old rdb file.
   It's dangerous for redis to delete the rdb file once persistence is disabled.

[sundb]

@singku In what scenarios do you need to frequently turn rdb on or off and restart it?

[fewtrell]

We're not so worried about folks doing frequent enable disable. We're more concerned about a one time disable & customer forgets about it. Then a long time later (maybe months?) they mysteriously load very out of date data into their database.

For a real world example if you have a larger number of users you will have some who are running very close to the memory or CPU limit for their system. Periodic snapshots may create too much memory or CPU pressure and the customer may choose to disable persistence in those cases.

[sundb]

The save config just turns off autosave, users can still generate rdb using save or bgsave commands, and it should be the user's responsibility to make sure they don't load old rdb files rather than letting redis automatically delete old snapshots, which would be very dangerous.
If you need to shut down the snapshot completely, you still need to change the dbfilename to avoid reading the old rdb or generating.

[singku]

It doesn't need to frequently turn on and off to trigger this issue. An one time turn off, and crash/OOM caused restart would lead to this "loading old data issue".

I am not debating on who should responsible for what work to make sure the system work correctly. But seeking for suggestions to see if it's reasonable to make this configurable. So for users, in stead of having to worry about "oh, to avoid loading stale data, I better delete snapshot files after disabling persistence". They can just configure the option on, then disabling would remove these files for users automatically.

[sundb]

it's easy to add this config, but it imposes higher learning costs and risks(data loss) on the users.
I would prefer that you use shell script that gets the save config in redis.conf, determines if it exists locally if it does, and deletes it if it does.