K8s POD delete data very slowly

[xiaoli007]

On the Ubuntu 24.04 operating system, with Ceph version 19.2.0, Ceph-CSI version 3.12.3, and K8s mounting PVCs via Ceph-CSI, performing an rm -rf operation on a 200GB file inside a pod takes about an hour. Without Ceph, on a standalone server, deleting 200GB takes around 10 minutes. It used Sata Hdd。 How can I optimize this issue?

[xiaoli007]

ceph pool used rbd mode

[nixpanic]

Ceph-CSI is not in the I/O path, it only creates/mounts RBD-images as volumes. There might be a configuration issue of some kind, but you would need to check with others that understand performance characteristics of Ceph better. Easiest is to reach out on their Slack or IRC and explain/ask there.