Question about RSS vs Cache memory usage in OpenBao/Vault

Vault
,
1

Hi everyone,

While running benchmarks on an OpenBao (Vault) cluster, I noticed an interesting memory behavior:

  • Initially, most of the memory was reported as cache (e.g., 7 GiB cache, ~200 MiB RSS).

  • After ~5 days of running, a large portion of that memory now also shows in RSS (5gb in RSS and 7GB in cache)

From what I understand, Linux uses free RAM for page cache, and applications like Vault/OpenBao will fault pages back into RSS when accessed again.

My questions:

  1. Is this expected behavior for Vault/OpenBao’s storage engines (raft)?

  2. Does Vault itself trigger compactions or maintenance tasks that could explain the shift from cache to RSS?

  3. Are there recommended ways to tune or monitor this (RSS vs Cache) so I can set appropriate memory requests/limits in Kubernetes?

Any guidance or similar experiences would be super helpful :folded_hands:

2

Hi @praneethpottimuthi, it sounds like you are using OpenBao and not Vault in relation to your question. This is a forum specifically for HashiCorp Vault. OpenBao is now its own project with its own implementation, so you will need to direct OpenBao-related questions to their community forums or issue tracking system. Thanks very much for your interest! If you have Vault-specific questions, please do use this forum to ask them!