Support Calxeda Highbank Nodes
The validation lab is in the process of deploying 2 Calxeda servers with 24 compute nodes each. These nodes support out-of-band management using IPMI and deployment with PXE. We need to add a "highbank" device implementation to the dispatcher.
<mwhudson> so installing a rootfs + "hwpack" (really just a kernel deb) might go like this:
<mwhudson> pxe boot a known-good kernel with an initrd that gives you a busybox shell on the console
<mwhudson> delete all partitions from /dev/sda
<mwhudson> create a boot and rootfs partition
<mwhudson> unpack the rootfs into the rootfs partition
<mwhudson> chroot into the rootfs partition
<mwhudson> mount the boot partition at /boot
<mwhudson> install the kernel deb
<mwhudson> reconfigure the node to boot off SATA
<mwhudson> restart the node
<mwhudson> ...
<mwhudson> profit!
<mwhudson> robher, fabo: sound basically sane?
<fabo> seems good to me
<mwhudson> and fedora will be similar except that "install the kernel deb" will have to be different
<mwhudson> ah heh, we will need to customize the master initrd a bit: needs fdisk, mkfs.* ...
Blueprint information
- Status:
- Complete
- Approver:
- Andy Doan
- Priority:
- Essential
- Drafter:
- Antonio Terceiro
- Direction:
- Approved
- Assignee:
- Nicholas Schutt
- Definition:
- Approved
- Series goal:
- Accepted for trunk
- Implementation:
-
Implemented
- Milestone target:
-
2013.04
- Started by
- Michael Hudson-Doyle
- Completed by
- Nicholas Schutt
Related branches
Related bugs
Sprints
Whiteboard
[doanac, 2013-01-29] this was blocked due to resourcing, but can now be picked back up for the 2013.02 cycle
[mwhudson, 2013-02-14] to summarize the state as of today: I think I know what needs to be done and how to do it, I "just" need to write the code and the documentation...
[danilo, 2013-03-08] Nick is picking this up with support from Antonio: mwhudson has handed it over.
[nicks, 2013-03-27] Tested deployment, filesystem access, and lava_test_shell through completion via the dispatcher using a temporary image based on ubuntu (not for production, just to allow internal lava development). Waiting on official working hwpack and snapshot to finalize implementation and to test it. Need to identify where to put documentation.
[danilo, 2013-03-27] Beta running with PXE and IPMI, but our image is not final (not based on hwpacks still). Code is not production quality so we can't roll it out yet. Bumping to next month to be completed.
[asac, 2013-04-24] Seems this was about getting calxeda codebase going and deploy it in staging; as of now I don't see it deployed in our production lab; where is that work tracked? Should this blueprint be renamed and acceptance adjusted to reflect that we delivered "beta" stage code and staging deployment of highbank ? Where do we track the production deployment? Maybe we can still do it today or tomorrow to deliver this month?
Meta:
Headline: Calxeda EnergyCore Support in LAVA
Acceptance: You can submit a LAVA job that runs on an EnergyCore node. LAVA can configure PXE boot with IPMI.
Roadmap Id: CARD-255
Work Items
Work items:
read calxeda documentation: DONE
boot server by hand: DONE
learn how to create custom initrd with extra commands: DONE
create 'master image' initrd: DONE
decide how the OS image will be specified in the job: DONE
implement support for deploying an rootfs & hwpack: DONE
implement support for accessing the filesystem after deployment: DONE
run test job through scheduler: DONE
write masses of documentation (how to set up dnsmasq, how to set up the device, etc): DONE
[terceiro] Notify Bernard Park when there is a working prototype: DONE
Dependency tree
* Blueprints in grey have been implemented.
