Re: [xcat-user] Discovery race condition ?

Er Tao Zhao Fri, 22 Mar 2019 07:21:09 -0700

Hi, Thomas

For "0c:c4:7a:4d:85:a8|0c:c4:7a:58:c7:6a!*NOIP*", so "0c:c4:7a:58:c7:6a" is the MAC address of eth2, right?

Will you pls show me the result of "tabdump discoverydata | grep 0c:c4:7a:58:c7:6a -A20"?

Thx!

Best Regards,
-----------------------------------
Zhao Er Tao

IBM China System and Technology Laboratory, Beijing
Tel:(86-10)82450485
Email: erta...@cn.ibm.com
Address: 1/F, 28 Building,ZhongGuanCun Software Park,
No.8 DongBeiWang West Road, Haidian District,
Beijing, 100193, P.R.China

----- Original message -----
From: Thomas HUMMEL <thomas.hum...@pasteur.fr>
To: xcat-user@lists.sourceforge.net
Cc:
Subject: Re: [xcat-user] Discovery race condition ?
Date: Thu, Mar 21, 2019 2:19 AM

On 2/25/19 10:04 AM, Er Tao Zhao wrote:
> Hi, Thomas
> Yeah, I agree with Kevin that if you not plan to bond eth0 and eth2,
> you'd better set then in different VLAN.
> But if eth0 and eth2 are in same vlan, xCAT can deal with it after xCAT 2.13
> Will you pls show me the output of `lsdef tars-113`
> And can `tars-113-eth0` or `tars-113-eth2` can be resolved to the same
> ip with `tars-113` in your DNS? And can be get from MN?
> For example `ping tars-113-eth0` or `ping tars-113-eth2` can get same ip
> in node `xcat-tars`
Hello,

as I said, sorry for the delay I've been in vacation then busy.

A quick reminder :

Nodes are normally physically configured like this :

conf 1:

a) eth0 -> switch A (1G) : ipmi (ipmi subnet) traffic allowed
b) eth2 -> switch B (10G) : only data (data subnet) traffic allowed

and logically configured to get switch-based discovered using swith B

I did configure on purpose one node/port like this :

conf 2:

a) eth0 -> switch A (1G) : data (data subnet) + ipmi (ipmi subnet)
traffic allowed
b) eth2 -> switch B (10G) : data (data subnet) only traffic allowed

precisely to be able to handle (by knowing what's happening and/or
forcing eth2 install) a switch misconfiguration which would result to
the conf2 above [and at the time I wrote my initial message, a BIOS boot
order misconfiguration too, but I'm not so sure it has something to do
with the MAC address ending up in the mac table]

What I'm seeing on the console is basically, once loaded and having got
an ip address on the BOOTIF nic ('Aquiring network address message'),
genesis getting 2 ips from the dynamic range by DHCP, one for each nic
after the 'Beginin node discovery process' message and the final MAC
registration in the mac table beeing for the eth0 nic despite the fact that

- discoverydata entry for tars-113 was cleared beforehand
- switch-based discovery attributes where those of eth2 (switch B)

So the node gets finally netbooted via eth0, which is not what I'd want.

- lsdef tars-113 beforehand was :

---
# <xCAT data object stanza file>

tars-113:
   objtype=node
   addkcmdline=ipv6.disable=1 biosdevname=0 net.ifnames=0
rd.driver.blacklist=nouveau nouveau.modeset=0
   arch=x86_64
   bmc=10.6.96.115
   bmcpassword=XXX
   bmcport=0
   bmcusername=XXXX

chain=runcmd=bmcsetup,runimage=https://urldefense.proofpoint.com/v2/url?u=http-3A__xcat-2Dtars_install_sum-5Factivate_sum-5Factivate.tgz-2Cosimage-3Dcentos6.10-2Dx86-5F64-2Dnetboot-2Dcompute-2Dprod&d=DwIGaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=GXZJILRYuh625D1P_w_8vxyNDAEGDWoFmPVwQ0NBUcY&m=R1A2OwgmdSDPRQP3MXYSuzUCIWGpAWk8bpM5NoHWq3Q&s=UJEv9keKLIZgJb82yOkFWl-SO8dvtOLom0qkSIJ_DtM&e=
   groups=tars-compute,tars-ipmi,tars,standard,b10
   ip=192.168.128.115
   mgt=ipmi
   os=centos6.10
   profile="">   provmethod=centos6.10-x86_64-netboot-compute-prod
   supportedarchs=x86,x86_64
   switch=b10b4.dc1.pasteur.fr
   switchport=8

- lsdef afterwards is

Object name: tars-113
   addkcmdline=ipv6.disable=1 biosdevname=0 net.ifnames=0
rd.driver.blacklist=nouveau nouveau.modeset=0
   arch=x86_64
   bmc=10.6.96.115
   bmcpassword=XXXX
   bmcport=0
   bmcusername=XXXX
   chain=runcmd=bmcsetup,osimage=centos6.10-x86_64-netboot-compute-prod
   cpucount=12
   cputype=Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz
   currstate=netboot centos6.10-x86_64-compute
   disksize=sda:256GB
   groups=tars-compute,tars-ipmi,tars,standard,b10

initrd=xcat/osimage/centos6.10-x86_64-netboot-compute-prod/initrd-stateless.gz
   ip=192.168.128.115

kcmdline=imgurl=https://urldefense.proofpoint.com/v2/url?u=http-3A__-21myipfn-21-3A80__install_netboot_centos6.10_x86-5F64_compute_prod_rootimg.gz&d=DwIGaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=GXZJILRYuh625D1P_w_8vxyNDAEGDWoFmPVwQ0NBUcY&m=R1A2OwgmdSDPRQP3MXYSuzUCIWGpAWk8bpM5NoHWq3Q&s=yfck6JfRE_2gEtvh0FAJDU-dFXebaciN9sc4OvYC27Y&e=
XCAT=!myipfn!:3001 NODE=tars-113 FC=0
   kernel=xcat/osimage/centos6.10-x86_64-netboot-compute-prod/kernel
   mac=0c:c4:7a:4d:85:a8|0c:c4:7a:58:c7:6a!*NOIP*
   memory=258373MB
   mgt=ipmi
   netboot=xnba
   os=centos6.10
   postbootscripts=otherpkgs
   profile="">   provmethod=centos6.10-x86_64-netboot-compute-prod
   serial=E162178X5A02118
   status=booted
   statustime=03-20-2019 17:55:02
   supportedarchs=x86,x86_64
   switch=b10b4.dc1.pasteur.fr
   switchport=8

- neither tars-113-eth0 nor tars-113-eth2 (fully qualified or not) can
be resolved

Again, the actual need is not to have 2 nics on the same subnet but to
have some way to choose which MAC will get discovered if so.

Does BIOS PXE order makes any difference to what nic gets into the mac
address at the end ?
Which nic does genesis picks up between the 2 to put in the mac address ?
Does a discovery happen on each nic ?

It seems like genesis actually did the switch-based discovery but
registered the first MAC it saw, thus the mismatch

Thanks.

--
Thomas H.

_______________________________________________
xCAT-user mailing list
xCAT-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xcat-user

_______________________________________________
xCAT-user mailing list
xCAT-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xcat-user

Re: [xcat-user] Discovery race condition ?

Reply via email to