Support for reloading configuration file?

[agittins]

In an environment where new databases are created at runtime, it would be helpful to have a way to ask litestream to re-read the configuration file.

We currently regenerate the config file and then restart the docker container that litestream is running in (we're using the official docker image). This works but I am wondering if it will cause a lot of new snapshots to be done on startup if our app happens to flush any WALs in the interim.

Does or can litestream support being sent say a SIGHUP to force reloading of the config file? (perhaps only loading new entries but not changes, if that's easier to solve)

[benbjohnson]

This works but I am wondering if it will cause a lot of new snapshots to be done on startup if our app happens to flush any WALs in the interim.

There's a window of time where Litestream releases its read-only transaction that blocks the checkpointing where your application could try to issue a checkpoint instead. In that case, Litestream will re-snapshot when it comes back up, however, typically this shouldn't cause new snapshots on restart.

Does or can litestream support being sent say a SIGHUP to force reloading of the config file?

A SIGHUP to reload would be a feature to add. Issue #42 is planned as well to monitor a directory (or matching glob) to automatically replicate new databases which would probably solve your problem better.

[agittins]

Thanks for the prompt response!

Yes, the globbing feature would probably work well for us, I've subscribed to the issue. We'll see how things go during testing - I think our app is unlikely to checkpoint any given db in the short time it takes for a restart so you are probably right that the impact will be low.