Hacker News Reader

Maybe I missed something. How do you account for the VM downtime? The VM containing the service writing to the sqlite db, it can go down without warning. How do you mitigate that?

View on HN

It doesn’t. Given that it periodically dumps the WAL to an S3 bucket for actual storage this sounds like async replication. Async replication is prone to lose last bits (last by time) of data when the node goes down. They’re probably more or less fine if the SQLite file is stored on the EBS though, it means they’ll just have to restart that node that failed or reattach EBS volume to a new healthy node. It does mean downtime though.

On another note I’m using Tailscale and it’s a wonderful service and I get it for free.

Reply on news.ycombinator.com