-
Most of the relocation time is spent to stop/start compute endpoints. Lets avoid that.
We can still detach tenant from pageserver compute is being connected to. (I think this is fine, need to discuss…
-
### Bug description
```
> [build 14/18] RUN npx prisma db pull:
1.179 Prisma schema loaded from prisma/schema.prisma
1.187 Datasource "db": PostgreSQL database "neondb", schema "public" at "som…
-
## Rationale
There's a few reasons:
1. The values "VM usage, else pod resources.requests" are used by cluster-autoscaler, and it [currently doesn't expose those metrics](https://github.com/kuber…
-
I run a production like load in staging with about 10000 projects created. All projects are deployed on the same page server.
On average 130 pgbench compute clients are running and 100 compute client…
-
We have a `/debug_dump` HTTP endpoint in safekeepers, to collect debug info about all timelines. It produces a large JSON object, which is currently held in memory.
Current implementation is here: …
-
## Steps to reproduce
Not entirely sure about that. But in logs It can be seen that in both cases there is this sequence:
1. create timeline call that times out and continued in the background w…
-
We have API in the pageserver for timestamp to LSN conversion, but the way it works could be a bit misleading. See this discussion, for example: https://neondb.slack.com/archives/C04382GUYCC/p16740441…
-
By default there is no branching available to users. So, someone who wants to try with the bare bones into NeonDB will struggle to find things without branch. If default purpose is not allowing branch…
-
The `/sql` interface of the http proxy doesn't handle an array of json documents as query parameter.
A plain json/jsonb query parameter works, and an array of any other datatypes also works, but an…
-
## Steps to reproduce
Create a new project, and run the following query:
```sql
DO $$
DECLARE
r RECORD;
BEGIN
FOR r IN (SELECT * FROM pg_available_extensions WHERE name NOT IN ('adm…