Martin Braun

https://docs.gunicorn.org/en/stable/settings.html#max-requests Reason: Memory usage is growing after a while for some customers. No memoryleak is known, this should be fixable via worker recycling though.

compose_client version negotiation

1

We need a way to do version negotiation between a) the server and the client and b) the dumped files and the client. Otherwise, the client and the server need...

The current cli library structure is a bit unweildy as we need to import from all kinds of different packages all the time. Instead we should come up with a...

FIXME: join the RETURNING * with the original values to construct the actual historical value

1

see comment here https://github.com/neuroforgede/nfcompose/blob/8ecd994ae3e5ae5f149cd8a4ef1a4219483d7125/skipper/skipper/dataseries/storage/dynamic_sql/queries/modification_materialized/insert.py#L313C10-L313C10

Should POST to datapoint (no upserts) throw errors on concurrent modification

1

Currently, the last POST silently wins when two start at a time where no entry exists because we are simply using the standard upsert logic. While this is good for...

see https://github.com/neuroforgede/nfcompose/blob/main/skipper/skipper/celery.py Some of these might take a while and we dont want them to pile up.

multipart uploads via axios (node.js) only work if you go through the proxy

1

this is an edge case that only happens if you are in the same network as the gunicorn container and directly use the gunicorn container as the target of your...

Implement compose_cli client support for Tenant and global User management

1

- [ ] Add support for Tenant diff and creation - [ ] Add support for global User management (without diff though?) - in the same swoop we should probably...

With our old backends DYNAMIC_SQL_V1 and DYNAMIC_SQL_MATERIALIZED we still store data in a fully materialized manner. Both backends are deprecated as of 2.0.1 and should be removed completely in 2.1.0....

see https://www.citusdata.com/blog/2023/07/18/citus-12-schema-based-sharding-for-postgres/