If you use xfs (+`file_copy_method=CLONE`) you can do this with Postgres 18.

`CREATE DATABASE clankerdb TEMPLATE sourcedb STRATEGY=FILE_COPY;`.

But Ardent can be useful for many, because cloud providers uses heavily restricted Postgres. And many use Aurora, which doesn't event let configure the `log_line_prefix`.

Though if cloud providers add file_copy_method=CLONE compatible managed pg ...

ref: https://boringsql.com/posts/instant-database-clones/

Here's how I do it with Heroku. Are there some cloud services that don't have an equivalent?

  heroku pg:backups:capture --app x
  heroku pg:backups:download --app x
  pg_restore --verbose --clean --no-acl --no-owner -h localhost -U postgres -d y local_db_for_robots_etc.dump

This takes more than 6 seconds. I'm curious how they achieved that for arbitrary DBs!

We've got docs on how we did it :)

https://docs.tryardent.com/architecture

But essentially we get around the restrictions of the original DB by replicating into a different postgres compatible DB that essentially serves as a read replica. That DB is the one that branches but since it mirrors the original DB you get effective clones

By doing this we get a lot more control over what we can do to create the clones. The read replica clones using copy on write + isolated autoscaling compute to clone in 6s. We use neon to do this since we think they've implemented those two properties well.

Since it's default postgres logical replication + DDL triggers you can technically point it at any "branching enabled" db on the other end in order to achieve the same effect

I wanted to try doing something similar to this in our dev environment (think shared dev database but per branch clones), but this limitation seemed tricky to accept:

> The source database can't have any active connections during cloning.

I wouldn't mind some lock contention, but having to kill all connections seemed a bit harsh

I guess it's still possible with a replica that no one is using?

So you don't need to touch real production db?

Not sure if it applies for all use cases tou

Yeah I think a read replica might fit the bill - though I suspect active logical replication counts as a connection in this context.

Using a cloud provider read replica might not (as I think that might use block level replication) - but then you're paying for an extra dev database host for the privilege

A little slow but on Aurora you can attach then promote read replicas. Iirc that's around 20 minutes but I haven't tested recently.

I'd think you could also setup logical rep to a VM then snapshot and clone the storage which is generally pretty fast.

You can create a new instance directly on AWS aurora. Takes less than 20 minutes!

  aws rds restore-db-cluster-to-point-in-time \
      --source-db-cluster-identifier <source-cluster> \
      --db-cluster-identifier <new-cluster> \                                         
      --restore-type copy-on-write \                                                   
      --use-latest-restorable-time \                                                   
      --db-subnet-group-name <sub group> \                     
      --vpc-security-group-ids <security group> \            
      --serverless-v2-scaling-configuration MinCapacity=0,MaxCapacity=16

Any idea on timing? I was under the impression pretty much anything RDS takes ~20 minutes although I haven't done any benchmarking/timing on the latest Aurora stuff

Main exceptions are copying/encrypting snapshots when they're not incremental and PITR restores which can take substantially longer.

Oh nice, the `/var` part of my workstation is a dedicated nvme drive and it's coincidentally formatted as xfs.

Companies like Ubicloud gives hosted actions faster and far more cheaper (5-10x) than Microsoft itself.

Now Microsoft will charge "data plane usage" (CRUDing a row that contains (id, ts, state_enum, acc_id ...) in essence) 2.5 more than what Ubicloud offers for WHOLE compute. Also to have "fair pricing" they'll make you pay 2.5 more the compute's price for being able to use their data plane.

cool.

it's rather egregious that it is a "per minute" tax rather than a $0.002 per job.

`new issue title" && curl hacker.boi -d $GITHUB_TOKEN && sleep 30` I hate pipelines even more :(