Pod request values can have mB unit when set as percentage of limit

[spadgett]

See memory usage below. Values like this also appear in the pod template for memory request. It looks like we're using mB to avoid losing precision when calculating request as a percentage of limit.

/cc @sspeiche @sosiouxme @jwforres @derekwaynecarr

[spadgett]

[derekwaynecarr]

A few options:

1. upstream could fail validation for any fractional request of memory
2. cluster overcommit admission controller can round up fractional values

I am pursuing 2 since I think 1 may impact backwards compatibility.

[derekwaynecarr]

Another thing:

the current code allows for cpu precision below a millicore. for example, 0.5m or 100.5m, we should also round those up.

[jwforres]

so derek's fix will get rid of the fact that there are mB but we will still
end up with B as the unit so...
@spadgett I discussed briefly with @smarterclayton and I do think we want
to humanize the values when we display them like you did with image size,
one decimal place of precision

On Thu, Apr 7, 2016 at 3:09 PM, Derek Carr [email protected] wrote:

A few options:

1. upstream could fail validation for any fractional request of memory
2. cluster overcommit admission controller can round up fractional values

I am pursuing 2 since I think 1 may impact backwards compatibility.

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#8391 (comment)

[derekwaynecarr]

@danmcp can this be labeled p1 since we are actually storing invalid data (i.e. a fractional byte, and a fractional millicore)