Are there Performance Metrics for the CEPH virtual-SAN IO-Wait Times?

[jorgw]

Hello,

Where please in CEPH are the io-wait times located? We think we need to measure those now since our 62TB are getting busy.


We have the VE 6.3-6 deployed on four physical Dell systems with all SSD drives, but our 50+ virtual-machines are now seemingly getting slower when disk-access is needed. The simple DataCenter view under CEPH-Performance only shows "Reads" and "Writes" and "IOPS Reads" and "IOPS Writes". No quick/simple view of queue lenght or wait-times there.

Maybe there is some documentation tucked away for those details and metrics please, so we can see if our 52 OSDs with their 512 PGs are all running at its best. Or maybe, since this is our first real Proxmox-Cluster, we need to learn and improve of what we did wrong. (also, all colors are nicely green and our ceph-version is displayed as 14.2.18 currently)


Thank you very much.


PS: Wouw, nice to see that a new version is available today. (VE 6.4 to be precise)

 

[aaron]

The performance stats shown by PVE are there to get a basic idea of how the system is performing. If you need very detailed information you will need to set up some external performance monitoring.

Ceph itself is able to talk to a variety of monitoring systems as well via the Manager. See the Ceph docs: https://docs.ceph.com/en/octopus/mgr/