3 ms·

Congratulations with your release! What is the current support level for `scale-out` clusters? I can find a thing or two about high availability [1] and a smal

Congratulations with your release! What is the current support level for `scale-out` clusters? I can find a thing or two about high availability [1] and a small set of transaction isolation levels [2], but I'm unable to find detailed information about locking mechanisms and cluster propagation. Is this a documentation problem, is it fully analogous to PostgreSQL or rather something that is still on the roadmap? [1]: https://www.edgedb.com/docs/reference/backend_ha https://www.edgedb.com/docs/reference/backend_ha [2]: https://www.edgedb.com/docs/stdlib/sys#type::sys::TransactionIsolation https://www.edgedb.com/docs/stdlib/sys#type::sys::Transactio...

RedCrowbar 4y ago
That `sys::TransactionIsolation` doc is outdated, EdgeDB only supports the `Serializable` transaction isolation. See related discussion [1] At this point EdgeDB supports Postgres HA passively, i.e. it will react to a failover event in your cluster via one of the documented mechanisms. Support for "active" cluster management is a planned feature too. Finally, EdgeDB server itself is fully stateless and you can run multiple instances of it in front of the same Postgres cluster. Admittedly, we need to document this better. [1] https://github.com/edgedb/edgedb/discussions/3466 https://github.com/edgedb/edgedb/discussions/3466
- Koffiepoeder 4y ago
  Thanks for the answer! It was clear to me that you were using Postgres under the hood, but it was not clear to me that your implementation was completely stateless. As a suggestions, maybe it might be useful to have a documentation page with some general EdgeDB architectural/internal structure information? I think that could be useful for future contributors as well. Anyway, for others looking into a completely separate deployment, running EdgeDB against a postgres cluster can apparently easily be done using the `EDGEDB_SERVER_BACKEND_DSN` variable in their docker container. https://www.edgedb.com/docs/guides/deployment/docker#edgedb-server-backend-dsn https://www.edgedb.com/docs/guides/deployment/docker#edgedb-...

RedCrowbar 4y ago

That `sys::TransactionIsolation` doc is outdated, EdgeDB only supports the `Serializable` transaction isolation. See related discussion [1] At this point EdgeDB supports Postgres HA passively, i.e. it will react to a failover event in your cluster via one of the documented mechanisms. Support for "active" cluster management is a planned feature too. Finally, EdgeDB server itself is fully stateless and you can run multiple instances of it in front of the same Postgres cluster. Admittedly, we need to document this better. [1] https://github.com/edgedb/edgedb/discussions/3466 https://github.com/edgedb/edgedb/discussions/3466

Koffiepoeder 4y ago
Thanks for the answer! It was clear to me that you were using Postgres under the hood, but it was not clear to me that your implementation was completely stateless. As a suggestions, maybe it might be useful to have a documentation page with some general EdgeDB architectural/internal structure information? I think that could be useful for future contributors as well. Anyway, for others looking into a completely separate deployment, running EdgeDB against a postgres cluster can apparently easily be done using the `EDGEDB_SERVER_BACKEND_DSN` variable in their docker container. https://www.edgedb.com/docs/guides/deployment/docker#edgedb-server-backend-dsn https://www.edgedb.com/docs/guides/deployment/docker#edgedb-...