Ticket #432 (closed task: fixed)

Opened 4 years ago

Last modified 4 years ago

CIENA-Ottawa reconfigure baremetal nodes

Reported by: ibaldin Owned by: vjo
Priority: major Milestone:
Component: Infrastructure: ExoGENI Racks OpenStack or xCAT Version: baseline
Keywords: Cc: jonmills, vjo, yxin


Marc L. wants to

1. Move some of the servers currently used for VMs to baremetal status

Q: What is the status of baremetal nodes booting at CIENA-Ottawa. Did we ever make a working xcat baremetal image?

2. Add two or three more new servers (with a different hardware configuration from the existing ones) to baremetal status

Q: Marc sent the specs of the servers on the list. Need to determine whether we need a separate class of baremetal servers (more work) or can fit them into existing class (less work)?

Change History

Changed 4 years ago by ibaldin

From Jonathan

I've configured the following nodes to be bare metal:


However, two issues remain:

1) I need a switchport change on the Ciena 5150 management switch. The ports to which ciena-w[4-8] connect need to be configured equivalently to:

switchport mode access
switchport access vlan 1008
and please ensure that portfast is enabled for these edge ports.

2) ORCA will need to be reconfigured to know something about them.

Changed 4 years ago by ibaldin

From Marc

Which port do you need to have access to Vlan 1008? em1 em2 or the 10GE port?

I figured it was em1 so it is added now for worker 4-8.

Changed 4 years ago by ibaldin

ciena-w8 is also a baremetal

Changed 4 years ago by ibaldin

As promised, works like a charm now:

[root@ciena-hn ~]# ssh ciena-w4?

[root@ciena-w4 ~]# cat /etc/redhat-release
CentOS release 6.7 (Final)

[root@ciena-w4 ~]# df -h
Filesystem Size Used Avail Use% Mounted on

6.4G 2.1G 4.3G 33% /

tmpfs 32G 0 32G 0% /dev/shm
tmpfs 10M 84K 10M 1% /tmp
tmpfs 10M 0 10M 0% /var/tmp

Changed 4 years ago by ibaldin

Still have a couple extra nodes with 40G. Need testing of the image.

Changed 4 years ago by ibaldin

From Jonathan:

Set up the em1 of the other nodes accordingly in the 5150 switch.

Its immaterial to me what you choose to do with the 40g interface in the 8700. However, it's possible I may need to alter the netboot image with correct drivers to support the 40g. I'll try to verify.

Changed 4 years ago by ibaldin

The 5 bm workers are now working. There are 3 more 40g capable nodes.

Changed 4 years ago by ibaldin

  • cc yxin added

For the additional 3 nodes the following RDF changes are needed:

1. Add ports 1/1, 2/1 and 2/2 for 3 additional bare-metal workers.

2. Make sure existing stitching vlans cover those ports as well

3. Define vlans 1779-1799 as 'NSI StitchPort?' (they will go to the NSI-controlled 8700 in StarLight?). These vlans should go to worker ports and port 4/1

Changed 4 years ago by ibaldin

Main issue I think is we need to be able to allow people to provision the three additional servers (they have 40G cards) as a separate resource class (they are different from the 5 other bare-metal nodes, which only have 10G ports). Not sure how much effort this is.

Changed 4 years ago by ibaldin

This could apply to OSF and SL racks as well, but less important, since they ONLY have 40G nodes.

Changed 4 years ago by ibaldin

I’ve checked in the code change and the new RDF for the 40G bare metal servers.

The config.xml needs to changed to add a new resource pool and delegation which for ciena rack, is cienavmsite.fourtygbaremetalce, number of 3.

Right now, the request needs to specify the new type in order to get it.

Please redeploy so I can test it in the rack. Forgot to mention, this is in the Master branch.

Changed 4 years ago by ibaldin

Latest version of Flukes (the production version) supports selecting 40G nodes in requests.

Still todo: xCAT handler has to be configurable to the node type somehow (regular baremetal vs. 40G). On Ciena rack there already are several node groups (baremetal-workers, 10g-workers and 40g-workers), so it's a matter of selecting appropriate group based on the resource type specified in config.xml

Changed 4 years ago by ibaldin

xCAT reconfigured to support 40G and tested.

Last task - verify NSI stitchport.

Changed 4 years ago by ibaldin

It looks like NSI stitchport is actually the InstaGeni? stitchport (vlans 1779-1790) in NDL, however it doesn't appear to work from local broker even though it has a delegation of cienaNet vlans.

'Embedding workflow error'

Yufeng - not urgent, but can you look in the next few days?

Changed 4 years ago by yxin

fixed the cienaNet.rdf and pushed i. please redelegate

Changed 4 years ago by ibaldin

  • status changed from new to closed
  • resolution set to fixed

This is now working. Closing

Note: See TracTickets for help on using tickets.