Slow CEPH Read/Write speed on cluster

I have experience with CEPH cluster. Both with Proxmox and with vanilla CentOS. Let me know if you are interested. Thanks!

Hi There, Which version of Proxmox you are using? Is this a new deployment? Since when you are facing low throughput? Can you share architecture diagram? Thanks,

Hey Hi I saw your screenshots. 1 - Your cluster is in ERR state and since you have posted a project I believe you have not tuned your cluster recovery to be fast enough to recover very soon in cases of HDD or Node failures. 2 - There are 5PG's for which continuous scrubbing is going on. Once it is done speed should improve. Again you need to do manual intervention to speed it up. 3 - Ceph has inbuild tool to check Read and write metrics, no need to use external tool. You are running say 20 VM's and testing the performance through just 1. How is that going to give you correct metrics. 4 - From image 3 it is clear that DEMAND of read / write is not huge, so a 5 node cluster should perform very well. 5 - I am not sure how you have configured the cluster but HDD based clusters should be backed by caching via SSD. It increases performance by 150%. Also if possible put WALDB on a separate partition and decrease your replica count. I have done Proxmox as well so I will be able to heal the cluster very soon. I do a lot of Ceph, Proxmox, OpenStack. Basically virtualization, cloud and Kubernetes are my core skills. I am limited by letters to brief my profile so more description on a call or something. Looing forward to do buisness together. You can reach me by replying back here or at IMSkype - singh.rahul.1603 Regards Rahul Singh