Sun Fire X2100 M2 – service processor woes (aka LOM)

Goodness me, the new service processor (aka LOM – Lights Out Management) in the Sun X2100 M2 is appalling! It sounds great in the marketing material, ie:
* ssh access
* serial cable access
* web gui access including java based KVM / remote desktop type thing
* once on SP you can connect to the system console, power off and on the box, etc etc

Well, the first two work, but then getting onto the system console seems to be highly problematic, and the java / browser based KVM thing doesn’t work at all for me, on any of the browser / OS combinations I have available to me (Mac OS X, Windows 2000; Firefox, IE, Safari). The system console access seems to freeze up shortly after the Solaris 10 install DVD gets going. If you plug in an LCD monitor you can see that it keeps going and starts up an X windows GUI solaris install, even if you’ve selected “serial console ttya” or “serial console ttyb”. The machine is deeply confused.

The simple task of installing a fresh Solaris 10 6/06 on this box is not being at all simple. Note, so far I’m refusing to plug a keyboard into it as that would be admitting defeat. The box is supposed to make it easy to install OS’s remotely, but this just isn’t working out in practice.

So I have a call in to Sun about it, as we have gold support with this box. They have reacted promptly but one of the first suggestions was to “upgrade the firmware” because “heaps of bugs” have been fixed since the version we have, including heaps of bugs to do with getting console access working.

Good lord, what use is a service processor if it can’t reliably give you console access to your running OS? Did they test this thing at all?

And so I go to the firmware download page and select the latest version of their firmware cd (1.4 at the time of writing, they seem to come out every few weeks!) and what am I greeted with but: “Sorry, we’re down for system maintenance…” Haven’t Sun heard of high availability? Or is this an example of their high availability systems at work?

FWIW, the various versions of things that shipped with this box are:

Embedded Lights Out Manager Version: 1.70
BIOS Version: S40_3A05

/SP -> show /SP/SystemInfo/BoardInfo

  /SP/SystemInfo/BoardInfo
    Targets:

    Properties:
        BoardManufacturer = Quanta
        BoardMfgDateTime = 2006/11/01 07.09
        BoardProductName = S40
        BoardSerialNumber = 2029QTF0638MB0348
        BoardPartNumber = 375-3343-01
        BiosVersion = S40_3A05

And what is with the user-unfriendly syntax? Whereas in the old ALOM and RSC service processors you could simplay say console to start a console session, you now have to go:

start /SP/AgentInfo/console

And instead of saying poweron you now have to go:

set /SP/SystemInfo/CtrlInfo PowerCtrl=on

etc etc.

Still, this syntax will be better for scripts to do things, and probably corresponds more nicely with an SNMP heirarchical breakdown.

La de da …

EDIT 2007-02-28 01:06

Well, turns out that you have to upgrade the firmware, and then if you want Solaris 10 update 2 (6/06) you need to workaround a bug by going into the BIOS (furiously hitting F2 in your SP console session during boot-up) and setting the IPMI External COM Port to System (instead of BMC), and then choose the Solaris 10 Serial Console ttya option on the install DVD. You then have to connect something to the serial ‘com 1′ port on the back of the server and use that to do the interactive install. Aparently if you use update 3 (11/06) you don’t have this restriction and things work nicer, ie you can ssh to the service processor and gain access to the console to do the interactive install that way.