Bug - The cluster total requested and memory limit are displayed in Ki
j-martin opened this issue · comments
Instead of Mi, Gi like every sub-totals per nodes.
...
memory 74249595.0Ki 64% 105203067.0Ki 90% 111.2Gi 10.9Gi
├─ gke-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 19562.8Mi 69% 29958.8Mi 105% 27.8Gi 0.0
...
Neat tool btw!
currently the "unit" for total is the lower unit the group to keep the accuracy, because the rounding caused more trouble (see #8). But I'm open to a better proposal.
Thanks.
My answer to #8 (comment)
I would round up to the first decimal.
1300 displayed as "1k" or "2k" ? => 1.3Ki
1500 displayed as "1k" or "2k" ? => 1.5Ki
1700 displayed as "1k" or "2k" ? => 1.7Ki
For example (above) 74249595.0Ki => 70.8Gi.
To me having a super accurate sum is less important than a more human-readable one. I am assuming the summations are done in bytes and the human-readable unit formatted when rendering the table.
If I read #8 properly, would it make sense to have a different unit "formatter" for CPU and memory (to avoid the said issue)?
Would it make sense to add a flag to force the sums to be displayed with the biggest reasonable unit for the table output? Arguably it may not be necessary because I would assume that anybody that cares about precision for automation (? or something else) would use the CSV output which uses bytes (as it should).
Hi,
The version 0.9.1 (from today) adjust scale/unit on every columns with the default display (non csv).
close the ticket if ok for you
Confirming the "bug" is fixed. Thanks!