apache / celeborn

Apache Celeborn is an elastic and high-performance service for shuffle and spilled data.
https://celeborn.apache.org/
Apache License 2.0
893 stars 361 forks source link

[CELEBORN-1706] Use bytes(IEC) unit instead of bytes(SI) for size related metrics in prometheus dashboard #2896

Closed turboFei closed 1 week ago

turboFei commented 2 weeks ago

What changes were proposed in this pull request?

Use unit bytes(IEC)(decbytes, 1,024 bytes in a kibibyte ) for below 18 metrics(disk and memory related) instead of bytes(SI)(bytes, 1,000 bytes in a kilobyte).

Also apply for 6 jvm metrics

Why are the changes needed?

Some size related metrics use bytes(IEC) and some use bytes(SI).

image image

The main difference between bytes in the International System of Units (SI) and the International Electrotechnical Commission (IEC) is the number of bytes in a kilobyte: SI: 1,000 bytes in a kilobyte IEC: 1,024 bytes in a kibibyte

FYI: https://www.drupal.org/project/drupal/issues/1114538#:~:text=According%20to%20the%20SI%20standard,e.g.%20a%20stick%20of%20RAM.

https://github.com/apache/celeborn/blob/4545cdc401a5274966453fe997f18fce666836f3/assets/grafana/celeborn-dashboard.json#L5636-L5699

Does this PR introduce any user-facing change?

Yes, metrics unit changed.

How was this patch tested?

Not needed, we already use decbytes in the dashboard json.

turboFei commented 1 week ago

do we have the contribution guide for modifying the grafana dashboard template?

No, currently it is a little difficult to add a new metrics.

pan3793 commented 1 week ago

Thanks, merged to main