PR #41 break xNBA DHCP for certain hardware

[ruixingw]

We have some very cheap motherboards on compute nodes, which may not be that mainstream (HUANANZHI X99 F8D, specs: http://www.huananzhi.com/html/1/184/185/491.html; the NIC is RTL8111H Gigabit Ethernet Card). We previously successfully provisioned OS using these motherboards with xCAT 2.16.1.

Today, I again got some of these machines and was trying to provision them with xCAT 2.16.2 and CentOS 8.4. However, it seems xNBA failed during DHCP, ending with a message like Configuring (net0 00:90:fb:6b:08:52).................. No configuration methods succeeded.
I also tried the command "dhcp" in xNBA(iPXE) shell. it doesn't work either.

Checking on the management node /var/log/messages, dhcpd received DHCPDISCOVER and sent DHCPOFFER, but these two repeat several times until timeout. There's no DHCPREQUEST nor DHCPACK at all. DHCP works fine for onboard PXE (which then chainloads xNBA I guess?) and other nodes.

Aug  6 17:22:15 master dhcpd[3315]: DHCPDISCOVER from 20:21:12:1a:03:d0 via eno2
Aug  6 17:22:16 master dhcpd[3315]: DHCPOFFER on 10.173.173.11 to 20:21:12:1a:03:d0 via eno2
Aug  6 17:22:16 master dhcpd[3315]: DHCPDISCOVER from 20:21:12:1a:03:d0 via eno2
Aug  6 17:22:16 master dhcpd[3315]: DHCPOFFER on 10.173.173.11 to 20:21:12:1a:03:d0 via eno2
Aug  6 17:22:18 master dhcpd[3315]: DHCPDISCOVER from 20:21:12:1a:03:d0 via eno2
Aug  6 17:22:18 master dhcpd[3315]: DHCPOFFER on 10.173.173.11 to 20:21:12:1a:03:d0 via eno2

Searching, I found PR #41 which upgraded xnba-undi from 1.0.3-131028 to 1.20.1-0 in xCAT 2.16.2. I downgraded xnba-undi to 1.0.3-131028, and the problem is solved.
xnba-undi-1.0.3 is well enough for us and I can just use it. I just want to let you know about this issue, just in case if there is any unseen bug elsewhere.

Let me know if you need more information or want me to do more tests.

[gurevichmark]

@ruixingw Have you tried xnba-undi-1.21.1-1 introduced by #47 and available here: https://xcat.org/files/xcat/repos/yum/devel/xcat-dep/xnba-undi-1.21.1-1.noarch.rpm

[ruixingw]

Thanks for the update. I'll try it when I get a chance

[besawn]

@ruixingw Have you had a chance to see if xnba-undi-1.21.1-1 resolves this issue yet?

[ruixingw]

@ruixingw Have you had a chance to see if xnba-undi-1.21.1-1 resolves this issue yet?

oops, this seems to be the last unresolved issue for 2.16.3 milestone.. I’ll try it asap (should be able to get it done in this week)

[ruixingw]

Unfortunately the problem still persists with xnba-undi-1.21.1-1 (see attached photo).

this new version does solve issue xcat2/xcat-core#7036 which I also encountered today with another hardware (Intel X520/DA-2, 82599ES controller).
since this current issue can be workarounded by downgrading to xnba-undi-1.0.3, I think it is okay to remove this from the 2.16.3 milestone and perhaps try to solve this in future releases.

[besawn]

@ruixingw Thanks for the feedback. I will move this issue to the next release milestone.