Why does Bitwarden self-hosted use so much RAM? And is it possible to tweak settings to make things better? (Proxmox VM)

Although the minimum recommended RAM is 2GB, my self-hosted installation is constantly using 2.8 to 3GB. It seems strange to me that software of this nature is always using that much memory, and I’m wondering if there’s any way to tweak things to make it run better?

I have Bitwarden running in a VM on a Proxmox server, and it uses more RAM than almost all of my other VMs and containers, which really doesn’t seem right to me.

The MSSQL server application fails to start unless it can allocate 2000 MB to itself.
At least, that’s what the logs said when I tried to run with a 2048 MB VM.

2560 MB was not enough, either. So I think the system requirements should be updated to say 3 GB instead.

I just opened an issue (on github on the server repo) on the docs not calling out enough ram. I am new to BW and misread MSSQL as MySQL and assumed that the RAM needs where correct, when I increased to 8Gig everything worked fine. The only issue is with the docs, no MSSQL should every be run on a system with less than 4Gig.

Use bitwarden_rs and be happy with a really low memory usage.

CONTAINER ID        NAME                      CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
0ff3xxxxx8        bitwarden.service         0.01%               6.168MiB / 985.9MiB   0.63%               174kB / 12.5MB      108MB / 184kB       15

I really like Bitwarden but this is one thing that bothers me. It would almost mean that I need a 4GB instance to self host, and that would mean around $20 a month instead of $5 small instances, which sounds like a hidden cost to self host, when Bitwarden cost is actually cheaper than that for a small team.

bitwarden_rs says it’s a “full implementation” but apparently not so full compared to the enterprise version.

When reading your complaint, I just had to like my own post above again: use bitwarden_rs for smaller installations, it consumes only a fraction of ram.
For larger installations the original package is recommended due to the more advanced database engine.