Closed msdisme closed 10 months ago
Plan is to move intel servers out of OCT 4 (ESI) rack and put 4 FX2's there. May be possible to get by with 4 or 6 u initially if necessary, instead of 8u. (so 2 or 3 FX2's initially) if we are overloading th MGHPCC teams
@CCI-MOC/okd @gagansk @hpdempsey may can you start with about 20 of the 32 servers and we will leave 10 for you and others to experiment with ESI? We can adjust up to the 32 as you need them, and as we add more servers from other locations to the pool
Hi @msdisme , Do you need anything from me to set these up?
Hi @derekhiggins, sorry I just spotted this query. I think the hardware is being added to the rack this week (we had a big donation of hardware which may impact that) and then we'll be reaching out to you to grab hold of them :-). OK if we start with about 20 nodes for your OKD work and put the other ones into ESI so that you and others may experiment with ESI?
Hi @derekhiggins, sorry I just spotted this query. I think the hardware is being added to the rack this week (we had a big donation of hardware which may impact that) and then we'll be reaching out to you to grab hold of them :-). OK if we start with about 20 nodes for your OKD work and put the other ones into ESI so that you and others may experiment with ESI?
Sounds good to me, just let me know when your ready so I can start.
Motivation
Why are we doing this? Enabling OpenShift upstream OKD will help ensure that the hardware predominantly used by NERC will be the basis for the upstream OKD. It also supports helping the larger community of open source as as a public good.
Completion criteria
Hardware installed in OCT 4 rack, networked to MOC networks and available for OKD team to use. OKD team performing regular nightly builds and metal based bare metail CI for current builds
Description
Changelog
2022- 10-18: updated with template 2022-11-07: added an epic link
Two projects,
Michelle strong emphasis on doing the POC to capture the current momentum and build on it.
Discussed some longer term options; eg. ESI or cloudlab to request servers and build on it. Emphasis on POC now, that as a mid to long term goal driven also by community.
Options for HW discussed for short term: The Need: 20-30 servers which they could have for at least 6 months (msd said let us plan on 12 months with a transition starting in 6 perhaps to ESI using them as a use case (share the ESI love with k8s !!))
My instinct is that we should go with 4 FC430's but need to discuss with team at BU.
Derek has some cool tools for pooling ci resources Ironic or equinix and add other providers - eg. ironic endpoint would work for ESI or provide feedback on improvements to ESI.
Epic