improve snapshot scheduling interface

[romange]

save_schedule flag is a bit confusing and not flexible enough.

Instead, we should introduce a more standard way of scheduling snapshots like cronjob expression.

this library can help with parsing and matching the events.
https://github.com/mariusbancila/croncpp

we should introduce a new flag (e.g. snapshot_cron?) and leave the old one to allow an easy transition before we deprecate it.

[eliskovets]

Hi @romange ,
Would it be possible to schedule saves only in case changes were made to the data?
In our case, we have a huge cluster that's mostly read-only, and we would like "save" to be triggered only if changes were made to the data.
The only possibility I see right now is to trigger bgsave when we finish writing data from the client code.

Thank you!

[kostasrim]

Hi @eliskovets,

Would it be possible to schedule saves only in case changes were made to the data?

Currently you can either save a snapshot explicitly via bgsave or periodically via the save_schedule flag.

Probably the best way would be to trigger it explicitly each time a client writes the data to the server.

Moreover, is there a specific reason why you would like snapshotting on write trigger?

[romange]

I agree with @kostasrim here. It's a very narrow use case and it requires from us some additional tracking and wiring to implement. Seems that calling BGSAVE/SAVE is an adequate solution for this problem.

[dor132]

@romange This can be marked as closed now.

[romange]

That was quick 😀. Thank you!

[romange]

I would like to apply some changes.

[dor132]

What are the changes?

[romange]

@dor132 I see you already saw - I think your proposal is good.