UCS

[Jonathan Bayless]

I have a Cisco UCS blade chassis system that we have just finished  setting up in the last few weeks. It is intended to run OpenStack and so  we are trying to install some variation of Ubuntu on it. Except that  when the blades boot, they seem to have a pretty big bug that is hitting  right after the network starts up. I have 4 UCS B200 M3 blades with 196GB of RAM and XEON E5-2620 CPUs with the VIC 1240 MLOM-40G-01 on each blade. They connect to UCS 6120XP FI switches. The hardware is all in 'operable' status and rebooting doesn't seem to make any difference.

I can finish an install of Ubuntu (12.04 LTS and 13.10 tested, both 64bit server installs) just fine, but on boot (about 5/7 boot ups) it get the following hang:

It has 4 vNIC network interfaces configured on each blade via a VIC1240  network board on the blade.

This  works fine with CentOS 6.5 x64 and Windows 2008R2 and 2012 server installs but Ubuntu hangs there,  gets an eventual DHCP timeout on occasion (because not all 4 vNICs are  on network VLANs that have DHCP running) and just sits there. I can hit  enter and it drops down a line, ctrl-C and it prints the ^c but nothing  else. alt-F2 or ctrl-alt-F2 whatever to try and spawn a new shell does  nothing. And I have let it sit for about 6 hours with no change.

On  rare occasions I have seen it finish boot, but it doesn't get that aepi  error on those occasions and seems to work fine. I was able to do an  apt-get update/upgrade and it completed everything without issue. The  network seemed fine but on reboot it went back to the same problem.

I have tried Ubuntu on 3 of the 4 blades I have available and have the same problem on each of them.

This is similar to https://supportforums.cisco.com/message/4123558?referring_site=bss&channel=bdp#4123558 but that one is mostly cosmetic. It causes a log entry but boots fine.

This issue prevents my servers from successfully booting about 75% of the time.

[Walter Dey · Accepted solution]

Don't know the solution, but

1) as of very shortly, Cisco officially supports Ubuntu, therefore you might open a TAC case

2) I have a customer, that successfully (after some troubles) installed Ubuntu on C-Series servers

Is it correct, that you install on local disk ? Raid-0, Raid-1

[Jonathan Bayless]

Where I have the UCS system up and running now, the blades have no local disk storage. They are booting from iSCSI with the use of a vNIC with the iSCSI overlay virtual adapter. This works great with windows and Centos. Ubuntu is the only problem.

I don't have support on this system (the boss is investigating it) so I can't open a case yet.

I have a second UCS system that I am setting up in the next day or two that will have local disks on the blades, so I will be trying Ubuntu again that way to determine if this issue is unique to iSCSI booting.

[Walter Dey · Accepted solution]

according to the latest interop matrix, Ubuntu is supported

see http://www.cisco.com/en/US/docs/unified_computing/ucs/interoperability/matrix/r_hcl_B_rel2.21.pdf

However, iSCSI boot is not supported.

This doesn't mean it doesn't work.

[Jonathan Bayless]

Ah I see, redhat is listed with a note saying iscsi boot is supported, but Ubuntu doesn't have that note. As you point out, there is nothing saying it is not supported which I assume just means it is not tested and proven supported. So yeah, basically I probably have to wait on Ubuntu for a while. That's too bad, it is definitely the more 'aggressively' supported platform for OpenStack (which is the whole reason we are interested in UCS. The ability to run everything, top down, through XML or similar API interface access and have templated and programmatically controlled datacenter integration is very attractive.

Hopefully Centos will be catching up with Ubuntu on its OpenStack support and integration shortly.