Add support for additional object storages like R2 or minio

[legout]

Description

There are a lot of (cheaper) alternatives to AWS S3, like Cloudflare R2, Backblaze, or self-hosted Minio. Support for additional object storages would be great. For example, in duckdb, or pythons fsspec there is a parameter to define the endpoint_url of the object storage.

[djouallah]

yes please specially R2 !!

[eitsupi]

Hi, I am trying to run a GlareDB server on CI. It would be great if we could also connect to a GCP emulator like https://github.com/fsouza/fake-gcs-server to make integration testing easier.

[YuriyGavrilov]

+1. could be useful with storj decentralized cheap storage.

[gruuya]

It would be great if we could also connect to a GCP emulator like https://github.com/fsouza/fake-gcs-server to make integration testing easier.

@eitsupi this is actually somewhat new and not yet documented adequately, but you should be able to do it if you build from source.

Here's how:

1. Spin up the fake gcs server; you can use a working example from here, this should result in a fake credentials file /tmp/fake-gcs-creds.json and a bucket (glaredb-test-bucket).
2. Spin up the server with a RPC bind, and configure it to use the fake gcs server using the location (-l) and option (-o) arguments:

   cargo run --bin glaredb -- server --disable-rpc-auth --rpc-bind 0.0.0.0:6555 -l gs://glaredb-test-bucket -o service_account_path=/tmp/fake-gcs-creds.json
   

3. To test it run a local glaredb instance and connect to the server

   $ cargo run --bin glaredb -- --ignore-rpc-auth
   Finished dev [unoptimized + debuginfo] target(s) in 0.69s
    Running `target/debug/glaredb --ignore-rpc-auth`
   GlareDB (v0.5.1)
   Using in-memory catalog
   Type \help for help.
   > \open http://localhost:6555
   Connected to remote GlareDB server: http://localhost:6555/
   > create table test as values (1, 'one'), (2, 'two');
   Table created
   > select * from test;
   ┌─────────┬─────────┐
   │ column1 │ column2 │
   │      ── │ ──      │
   │   Int64 │ Utf8    │
   ╞═════════╪═════════╡
   │       1 │ one     │
   │       2 │ two     │
   └─────────┴─────────┘
   

4. Verify the contents in the bucket

   $ curl -s --insecure http://0.0.0.0:4443/storage/v1/b/glaredb-test-bucket/o | jq .
   {
     "kind": "storage#objects",
     "items": [
       {
         "kind": "storage#object",
         "name": "databases/00000000-0000-0000-0000-000000000000/tables/20000/_delta_log/00000000000000000000.json",
         "id": "glaredb-test-bucket/databases/00000000-0000-0000-0000-000000000000/tables/20000/_delta_log/00000000000000000000.json",
         "bucket": "glaredb-test-bucket",
         "size": "1206",
         "crc32c": "3VCvgQ==",
         "md5Hash": "5IB4pt04RVTwx/jdycJc7A==",
         "etag": "\"5IB4pt04RVTwx/jdycJc7A==\"",
         "timeCreated": "2023-10-18T07:48:23.437901Z",
         "updated": "2023-10-18T07:48:23.437943Z",
         "generation": "1697615303438180"
       },
       ...
   

[tychoish]

@scsmithr just wanted to see where we're sitting with this one. My memory is that there's just documentation issues remaining....