Hey guys, I have a question regarding a deployment of LakeFS on on-prem machines. We have a POC ready deployment that was implemented with docker compose postgresql db, on the same server, and a minio S3 on another host. Is hosting the db and the lakefs on the same server is a good approach? Also, another question regarding memory, we see that the loading is slow on initial login, I saw an error regarding the GC rules and configured it, IDK if it really helps with the loading issue, but what is the best way to give more memory to the deployment and make the loading faster? As far as I know I should not do anything because docker takes care for the memory etc..? Thanks :)

Hey Yarin, welcome to the lake lakefs lakeFS uses the DB for storing some of the versioning metadata. Like any other online service, a highly available lakeFS cluster requires more than a single host running the lakeFS server. If your plan is to eventually run more than a single lakeFS server, than they all should be connected to the same DB as they share the same metadata. Assuming you’re ok with the availability of a single server, you can run the DB on the same server. Do note that any loss/corruption to the DB will result in inconsistencies and/or lakeFS being non-functional, so our recommendation is to adhere to DBs best-practices, like backups. The best way to debug the initial login slowness (assuming it’s UI) is through the network tools. There’s probably some request that is taking longer than it should. Given that, we could follow up on the needed adjustments to memory, object store, DB, etc.

Thanks!