Using volumes for committed cache: will it work if multiple pods shared the same directory for caching?

Good question. I think not. Each instance handles the local cache and takes locks when deleting/inserting a file. That lock is local.

I would expect it not to work. I am not sure the engineering is worth the gain. Generally the way to scale lakeFS should be to have larger rather than more pods. So pods should try to share no local resources, in the expectation that the production pods should have as much CPU and network bandwidth as possible. Even in future, when we remove PG entirely, I still expect this to hold.

Thanks!