Closed rata closed 7 years ago
Should we just default the resource to this?
On Tue, Oct 04, 2016 at 07:34:56AM -0700, Brandon Philips wrote:
Should we just default the resource to this?
Didn't look at the code to see if this is expected to remain true in bigger
clusters. Also, I don't know if we want to make such a promise now. Maybe it's
too soon and we might want get something slightly more expensive (1m
cpu is
really nothing).
But a note saying it's lightweight and an example of how small this can be for a small cluster, I think it helps to make the reader not hesistate on this (or not much).
When it's more mature, we may feel confident to do it, but I think now might be too soon (although I'm not involved with the code, maybe it's not too soon!). Or, when there is an update strategy, this might not even be needed :)
Thanks a lot, Rodrigo
@philips friendly ping? :)
Sure, LGTM
@squat fixed, thanks!
I am OK with this. I would be OK changing the default to 1m CPU request and 10m limit. Memory request just needs to be enough to not OOM.
My main concern to run this in my production k8s cluster was it's CPU/mem usage. So, for example, if one node dies I was afraid this pod might get scheduled and not others, more important, pods.
But, looking at it, it's really lightweight and can fit in any node just fine without causing any issues (nor having more nodes). So, it may be useful for other to see there is really no risk.