Failed to add compute node

[choonho]

When I add compute node by PXE. There is no compute node in the servers list.

And headnode says "Requires Setup". This must be compute node. The Network Interfaces of headnode is changed to compute node.

The correct information was

[askfongjojo]

@choonho, can you ssh into the headnode, execute the following, and let us know what is returned?

sdc-cnapi /servers

[choonho]

[root@headnode (CPG) ~]# sdc-cnapi /servers
HTTP/1.1 200 OK
Content-Type: application/json
Content-Length: 607
Date: Thu, 20 Oct 2016 05:23:40 GMT
Server: Compute Node API
x-request-id: 65c5202b-5175-472f-8f77-279484c2323a
x-response-time: 6
x-server-name: 0b0cf1c7-8d90-42eb-8532-dcf2bc421c0e
Connection: keep-alive

[
  {
    "datacenter": "CPG",
    "overprovision_ratio": 1,
    "reservation_ratio": 0.15,
    "reservoir": false,
    "traits": {},
    "rack_identifier": "",
    "comments": "",
    "uuid": "00020003-0004-0005-0006-000700080009",
    "reserved": false,
    "boot_platform": "20160929T030056Z",
    "boot_params": {
      "rabbitmq": "guest:guest:rabbitmq.CPG.mobile.local:5672"
    },
    "kernel_flags": {},
    "default_console": "serial",
    "serial": "ttyb",
    "created": "2016-10-20T01:27:04.000Z",
    "ram": 24567,
    "hostname": "headnode",
    "status": "running",
    "headnode": true,
    "current_platform": "20160929T030056Z",
    "setup": true,
    "last_boot": "2016-10-20T01:35:23.000Z",
    "transitional_status": "",
    "setting_up": false
  }
]

[askfongjojo]

The CN add seems unsuccessful. As you can see, there is only one server record and it is for the headnode. But somehow the setup appears to have rerun for the headnode, because AdminUI has found that the most recent setup job for this server uuid has failed and displays the "requires setup" prompt. During the course, it might have re-provisioned admin nics for it.

It'll be helpful to locate all the server related jobs and examine the sequence of events and where things have gone wrong.

[choonho]

The first error at Jobs is

napi.update_nics first of 1 error: {"message":"napi_nics::161331843469 was expected to have etag "BFEE8A62", but was "3A4F8765""}

and this may be related with

[choonho]

When I test sdc-agents-healthcheck, it shows

[root@headnode (CPG) ~]# sdc-agents-healthcheck 00020003-0004-0005-0006-000700080009 provisioner error 00020003-0004-0005-0006-000700080009 zonetracker error 00020003-0004-0005-0006-000700080009 ur ok 00020003-0004-0005-0006-000700080009 heartbeat error [root@headnode (CPG) ~]#

Is this problem?

[matthewpucc]

It looks like I'm running into this same problem. On a default install, new ixpe booted nodes are being deployed as a headnode. I've tried adding "headnode": false, to the kernel arguments. I also added a hostname of "node-init". The ipxe booted compute node comes up with this new hostname but the original headnode still see's it as the same guid and marks it as "headnode" as well.

Has there been any progress in tracking this down?

[bahamat]

This is happening because there are multiple CNs with the same UUID. This confuses CNAPI because it thinks they are the same CN. You'll need to contact your hardware vendor for a utility to update the SMBIOS UUID.