Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.
This project is now read‑only. Starting Monday, February 2, please use https://ibm-ceph.atlassian.net/ for all bug tracking management.

Bug 2370442

Summary: [GSS] Create an alternative and more reliable way to benchmark disk devices and get reliable MAX IOPs values.
Product: [Red Hat Storage] Red Hat Ceph Storage Reporter: Raimund Sacherer <rsachere>
Component: RADOSAssignee: Sridhar Seshasayee <sseshasa>
Status: CLOSED UPSTREAM QA Contact: Pawan <pdhiran>
Severity: high Docs Contact:
Priority: unspecified    
Version: 9.0CC: bhubbard, ceph-eng-bugs, cephqe-warriors, nojha, sseshasa, vumrao
Target Milestone: ---   
Target Release: 9.0z2   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2026-03-04 09:53:20 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Raimund Sacherer 2025-06-05 14:39:51 UTC 
Description of problem:

From Support we are not sure about having an OSD benchmark itself on startup.

- On node reboot, having all OSDs start up at the same time with all the other things the Operating System does might result in cpu starvation or the benchmark thread not getting enough BW or some other saturation. 
- Same situation actually when restarting an OSD, if there is so much cluster/client load going on that the benchmark would not get realistic data (backfills, recovery or just plain high CU load).

We where thinking that those benchmarks would be best taken on deployment or scale out, one disk at a time with no workload to make sure there is no interference and we can get the complete potential of a drive, would that make sense?



Engineering comment based on discussion in document linked below:
```
Sridhar: Yes, for tentacle and beyond, we want to change the way benchmark is performed on an OSD. One of the solutions being considered is to explore running the benchmark using fio or an equivalent tool on deployment. More on this when I have something more concrete.
```

BZ opened on suggestion from Sridhar to have the request tracked in BZ. 


Version-Release number of selected component (if applicable):
N/A


Additional info:

Mclock discussion about issues seen with Customers between Support and Engineering:
https://docs.google.com/document/d/1scCV9_mo-P6TsJZYCcJyceLEg29Y0GaR6YbIWQHsjdg/edit?tab=t.fxx5516gggj7
Comment 4 Red Hat Bugzilla 2026-03-04 09:53:20 UTC 
This product has been discontinued or is no longer tracked in Red Hat Bugzilla.