"...but I am sure that in their zeal to make computers into sealed undebuggable Windows bricks, Intel did not bother to include one."
Probably if Intel had intended the NUC to be a server running SmartOS there would be a serial port. They are not sealed, there is no OS (not even a SSD included), and they don't have to run Windows. Undebuggable? Not sure. Never had the need to because I use them for what they were intended to do and they work perfectly. Anyway, as Keith mentioned, SmartOS is not going to work on all hardware, and I'm glad it doesn't. For what it's worth, SmartOS also will not boot on one of my IBM low-end servers (IBM not Lenovo). Solaris 11 does, and so does OmniOS. So, I buy approved hardware and move on. Stability in my datacenter is important enough to make damn sure my hardware is on the HCL. On Mon, Nov 24, 2014 at 9:53 PM, Keith Wesolowski via smartos-discuss < [email protected]> wrote: > On Tue, Nov 25, 2014 at 07:57:53AM +0530, Sam M via smartos-discuss wrote: > > > I had posted this earlier, and there was quite a lot of feedback. Thanks > > all for this. So I should conclude that I will not be able to get the USB > > keyboard to work on the NUC DN2820FYKH? > > > > I'm just looking to solve the problem for now, and I can't believe > SmartOS > > not being able to handle a USB keyboard is the issue here. It's almost > > 2015! The Linux's have no problems with the NUC. > > USB keyboards work fine in SmartOS; they're what we use in our data > centres, labs, and repair depots for crash carts when the server's BMC > isn't working (i.e., no serial access). There's probably some problem > with the specific combination of USB controller, firmware, keyboard and > its firmware, and something in the USB software stack. > > To be fair, though, SmartOS is not a consumer operating system. The > target audience is data centres, and that means servers. The NUC is > about 1/32 the size of the smallest thing we'd deploy ourselves, and > targets a completely different market using a completely different > chipset. It shouldn't surprise anyone that we haven't made testing and > debugging firmware bugs on the NUC a priority. But you can, if it's > important to you! > > My first question would be whether you can get into kmdb (and actually > pass it input). If you can, you'll have a much better chance of > debugging this. If not, I'd start by figuring out which component needs > to be debugged by swapping things around with known-working systems > until the culprit (or at least trigger) is isolated. For example, if > you can boot the NUC with a different keyboard and use that to set up > networking, you could then replace that keyboard with the bad one and > start debugging. The specifics will depend on what works with what and > when. Everything in this process will be a lot easier if the NUC has a > real serial port you can use, but I am sure that in their zeal to make > computers into sealed undebuggable Windows bricks, Intel did not bother > to include one. > > > Is there any way I can automate the installation of SmartOS on the NUC, > so > > that I don't need a keyboard, by perhaps putting a script or JSON file on > > FAT partitioned sector of the boot disk? This file would answer the > > questions the installer usually asks, including specifying the IP > address. > > Once I have SSH access to the box, I'll be able to do the rest. > > SDC can do this; SmartOS generally cannot. However, I suspect that you > could modify > > https://github.com/joyent/smartos-overlay/blob/master/smartdc/bin/smartos_prompt_config.sh > (and build yourself a custom platform) to support such a feature. > > > Or can I SSH into the box after booting the box via the "noinstall" > option > > and then run the setup script? > > Not unless something else has somehow brought up networking. I don't > see how. Everything other than normal operation is built around the > assumption of access to the console device, which is normally not a > keyboard and monitor but a remote-serial or IPMI SoL device. Without > that, you're not going to get anywhere. > > > If this capability is not there in SmartOS, I think it may be worth > looking > > into. > > Agreed. Having automated setup is very useful in working on SDC, and I > am sure it would be useful in SmartOS too. Improvements to the setup > tools would be welcome. The two paths I'd consider here are: > > 1. Improving the above prompt_config bits independent of SDC. By > "improving", I mean "seriously consider burning to the ground". At > minimum, unattended installation and automation of disk selection and > zpool layout (using the disklayout(1m) and mkzpool(1m) tools in the > platform) are essential features. > > 2. Eliminating the separate SmartOS platform entirely by unifying the > SDC and SmartOS platform images, with a boot-time option similar to > noinstall to select SmartOS (vs. SDC HN or CN) operation. This is a lot > more work but it also deletes a lot of unloved code and gets everyone > onto a single well-maintained path. > > If you've never worked on this before, (1) is probably an easier place > to start; even if you eventually do (2), that small project will serve > as an introduction to the SmartOS platform build, the tools, and the > setup process itself. > > Let us know if you have questions. > > > ------------------------------------------- > smartos-discuss > Archives: https://www.listbox.com/member/archive/184463/=now > RSS Feed: > https://www.listbox.com/member/archive/rss/184463/23055384-4ad60763 > Modify Your Subscription: > https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com > -- Mark ------------------------------------------- smartos-discuss Archives: https://www.listbox.com/member/archive/184463/=now RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00 Modify Your Subscription: https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb Powered by Listbox: http://www.listbox.com
