Conrad Ludgate

> One alternative might be to speed up the PBKDF2 computation more directly with a cache. PBKDF2 is computed as: > > ``` > U1 = PRF(Password, Salt + INT_32_BE(i))...

> Would recommend logging failures per role or tracking usage to catch abuse Checked the code, and we already do log all failed authentication attempts, with the endpoint and role...

> LGTM, although initially was a bit confused with naming. Maybe just to add TODO in code to have clearer names for connect_ methods and functions? You're right. I didn't...

> I think we could share the cargo chef prepare layer with that. Sounds good. Unfortunately it seems the mounts we have set up are directly conflicting with chef, which...

I wrote the following test: ```rust use tokio::{ io::{AsyncReadExt, AsyncWriteExt}, net::{TcpListener, TcpStream}, time::Instant, try_join, }; #[tokio::test] async fn throughput() { let proxy_listener = TcpListener::bind("0.0.0.0:0").await.unwrap(); let compute_listener = TcpListener::bind("0.0.0.0:0").await.unwrap(); let proxy_addr...

Prior attempt: https://github.com/neondatabase/neon/pull/11765 Looks like it is used in e2e tests as well.

~~Seems to be bootlooping in testing. Moving back to draft~~ I was holding it wrong (using non vm image...)

~~Hmm. This doesn't yet work for accessing the aws-iam-token volume mount which is automatically created for us. I'll have to think more for a potential solution. Ideally we'd have some...

> @conradludgate may I ask to provide more context for the PR like what the issue is it related to and what do we want to achieve? We struggled a...

How strange. Do new playlists with public=false properly get set to private?