Clojurians Log v2

Just deploy a PostgreSQL somewhere and make sure that it is encrypted at rest and accessible only through HTTPS, possibly inside a VPC. If you are using checkpointing you need to deploy/configure storage, eg. S3.

> Just deploy a PostgreSQL somewhere Just to add that using Postgres for both tx-log and doc-store is generally okay, and definitely simplifies backup/restore, but the cost-performance tradeoffs can vary (e.g. it's going to be more expensive for certain workloads vs. just using S3 as the doc store) so it can be worth benchmarking I would also recommend not embedding XTDB into your app nodes too much - like that's how v1 was designed to be most flexible from a query perspective, but it makes it more complex from an ops perspective (app server restarts therefore restart the database also etc.) Out of interest, how come you're looking at v1 and not v2 here?

Ah yes, we use checkpoints to help us with app local RocksDB indices, thus redeploys do not reset the local index and we get decent startup times

As the app container disks are usually ephemereal

@taylor.jeremydavid My domain data (medicine/health) is graph based and type/schema flexible and doesn't really work nicely with the relational/tabular paradigm. I tried modelling the problem in Postgres before and it was a massive headache trying to force it into a crapton of recursive CTEs and views. Basically I needed a Datalog database because Datalog seems like it can express common clinical questions more concisely/elegantly (e.g. "how many COPD exacerbations requiring ICU admission has this patient had in the last year", "list all the antecedent causes of death ranked by precedence").