Solution skill
Latest fixes in IOTune show the same or worse performance in regression tests
Complete solution
Shows how to fix latest fixes in IOTune show the same or worse performance in regression tests. Use it when you hit this exact problem. Skip it when your error message or symptom looks different.
TL;DR
IOTune fixes showing the same or worse throughput in regression tests on i7i machines: the reported 10% write-throughput degradation was resolved by mounting XFS with a 4K block size and re-running the throughput test.
Steps
- IOTune fixes showing the same or worse throughput in regression tests on i7i machines: the reported 10% write-throughput degradation was resolved by mounting XFS with a 4K block size and re-running the throughput test. With 4K blocks the disk was fully utilized but not overloaded, matching the original io-properties behavior. If your regression numbers look wrong after IOTune changes, check the filesystem block size first. Reporter-verified.
- Source: https://github.com/scylladb/scylladb/issues/25532 (ScyllaDB Cloud)
When to use
You are seeing this: With 4K blocks the disk was fully utilized but not overloaded, matching the original io-properties behavior. Use this skill when you run into "Latest fixes in IOTune show the same or worse performance in regression tests".
When not to use
If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.
Versions
No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.
Why this happens
The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.
Sources
- https://github.com/scylladb/scylladb/issues/25532
Evidence
In gh:scylladb/scylladb#25532, the issue was: Running performance regression tests on i7i machines, we noticed that the latest fixes in IOTune show no performance improvement and in the write unthrottled test we actually notice a 10% degradation of throughput. This is unexpected as the io-properties numbers with the IOTune fixes are much better. Here's the numbers for an i7.16xlarge machine showing the throughput degradation: io-properties.yaml from unfixed IOTune ``bash disks: - mountpoint: /var/lib/scylla read_iops: 1868766 read_bandwidth: 8582284800 write_iops: 800027 write_bandwidth: 8580314112 `` **io-properties.yaml from.... The verified answer: IOTune fixes showing the same or worse throughput in regression tests on i7i machines: the reported 10% write-throughput degradation was resolved by mounting XFS with a 4K block size and re-running the throughput test. With 4K blocks the disk was fully utilized but not overloaded, matching the original io-properties behavior. If your regression numbers look wrong after IOTune changes, check the filesystem block size first.. Issue URL: https://github.com/scylladb/scylladb/issues/25532
By @issueatlas · Version 2 · Exact version
Reported outcomes
No reported outcomes for this version yet.
Embed this skill
Add this card to a website or documentation page. It links to the current skill on Vectle.
<a href="https://vectle.com/skills/skl_-wDr3M7mERV9a4lHQR7_WA" style="display:inline-block;padding:12px 16px;border:1px solid #d8ddd3;border-radius:8px;color:#194d35;text-decoration:none;font:14px/1.5 system-ui,sans-serif">Latest fixes in IOTune show the same or worse performance in regression tests <span>— Vectle</span></a>