Stop discouraging AOF

[zuiderkwast]

On the page about persistence under Ok, so what should I use?, we recommend using RDB or RDB+AOF but discourage AOF alone:

There are many users using AOF alone, but we discourage it since to have an RDB snapshot from time to time is a great idea for doing database backups, for faster restarts, and in the event of bugs in the AOF engine.

Is this really true? I think this is outdated and incorrect. Using AOF alone seems perfectly fine to me, and using an AOF file for backup should be fine (although a single RDB file might be simpler to handle). Using both seems like a lot of unnecessary forking, if the data is large.

@valkey-io/core-team WDYT?

[hwware]

Now we have a config aof-timestamp-enabled to record the timestamp, thus AOF is useful for restoring data to a specific time. We could remove the words "discourage"

[zuiderkwast]

AOF is useful for restoring data to a specific time

How to restore from a specific timestamp? Is this anywhere in the documentation?

[zuiderkwast]

and in the event of bugs in the AOF engine.

Actually this part seems to be true. There are bugs. There are fixes too. I think we shall not remove this disclaimer until we have at least fixed/merged these:

• Allow partial sync after loading AOF with preamble valkey#2366
• Skip AOF rewrite when a short write occurs valkey#2627
• config option to stop master on AOF write error valkey#2375
• add parameter for initiating bgrewriteaof on exceeding maximum AOF size valkey#2368

[madolson]

I don't think we should discourage the use of AOF, but I am in agreement I don't think it is the most robust mechanism today.

[soloestoy]

In terms of data recovery, AOF is clearly better than RDB because it has more real-time incremental data. We should not discourage on this usage, but rather make it more robust.