A RESTART checkpoint during a snapshot transfer locks the database for writing until the snapshot has finished

[t-voet]

Litestream v0.3.5 and a 25 GB database on ZFS with 64k page_size. A SFTP snapshot takes hours, during which the WAL hits 10k pages (640 MB). An obvious solution could be skipping RESTART checkpoints while a snapshot transfer is in progress.

[benbjohnson]

Thanks, @t-voet. I'll take a look at optimizing SFTP and skipping the RESTART checkpoints during transfer for v0.4.0.

[t-voet]

You're the best. If you send me your public key, I can provide a high-speed (2.5 Gbps) SFTP account for testing.

[hifi]

As I believe #477 fixed this I'll close. Thanks.