-
Notifications
You must be signed in to change notification settings - Fork 208
Eliminate FPGA admission webhook's mode #301
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
So, in this patch I made the FPGA plugin expose AFs as The problem though is that such resource name is 64 bytes long (32 + 32). Whereas the max resource name length without namespace is 63:
/cc @kad @bart0sh Do you mind if I remove the last character of FPGA interface IDs when exposing AF resources? |
I don't. At least for the POC version. |
Problem:
It is possible to run the FPGA device plugin in two different modes on different nodes of the same cluster. Yet the admission webhook can be aligned to work with FPGA device plugins in either preprogrammed or orchestrated mode. The webhook needs to be redesigned to be agnostic about the device plugins' modes. Also when operating in preprogrammed mode it is impossible to differentiate nodes providing the same accelerated function with different hardware, e.g. a user's request to dispatch a task onto
stratix10-dcp1.0-nlb0
may well be dispatched to a node running the nlb0 accelerated function on an Aria10.Solution:
-mode
option and to translate requested resources using AcceleratedFunction CRDs only in the new format.The format of resource names visible to a user is not changed. Basically the format can be anything, but it's expected to be in the form
<hardware>-<firmware_release>-<accelerated_function>
, e.g.arria10-dcp1.1-nlb0
.The text was updated successfully, but these errors were encountered: