Prev: Cannot boot x4200 via dhcp (from another x4200)
Next: Solaris 10 tape drive permission issues...
From: MrFluffy on 6 Mar 2007 07:24
On 24 Feb, 14:09, simoninusa2...(a)yahoo.co.uk wrote:
> I've just bought a Netra T1 105 and as it has no CD-ROM, I want toJumpstartit.
>
> I've tried setting up aJumpstartserver using Linux (Fedora5) and
> also from a Solaris 10 11/06 x86VMWareinstance, and both time the
> "boot net - install" fails at the same point, with a "38e00 |" prompt,
> here's the LOM output (serial/mac/hostid removed):
>
> lom>
> LOM event: power on
> .
> Netra t1 (UltraSPARC-IIi 440MHz), No Keyboard
> OpenBoot 3.10.27 ME, 1024 MB memory installed, Serial xxxxxx.
> Ethernet address xxxxxx, Host ID: xxxxxx.
>
> Drive not ready
> Boot device: sec-disk File and args:
> Drive not ready
> Evaluating: boot
>
> Can't open boot device
>
> ok boot net - install
> Boot device: /pci(a)1f,0/pci@1,1/network@1,1 File and args: - install
> 38e00 |
>
> Swapping out the drives from a D130 makes no difference, other than
> the "drive not ready" bit from LOM gets replaced with:
>
> "Boot device: pri-disk File and args:
> :217: The file just loaded does not appear to be executable."
>
> add_install_client seems to have worked properly (on Solaris/VMWare)
> and creates pretty much the same bootparams/dfstab/ethers/hosts/
> tftpboot entries as the Linux tutorials. It couldn't find adb, but the
> IP addresses all are fine. setup_install_server worked fine too.
>
> I've tried using sun4u and SUNW,UltraSPARC-IIi-Engine as the platform,
> no difference there.
>
> >From the output of tcpdump and the server's console, it would seem
>
> that the Netra is getting an IP (I can ping it) from rarpd and /etc/
> ethers, and a boot image from /tftpboot, but then it just seems to sit
> there when it gets to the NFS bit, here's some of the dump:
>
> 20:00:51.969246 IP 192.168.0.79.nfs > netra.157583: reply ok 80
> getattr
> 20:00:51.969407 IP 192.168.0.79.nfs > netra.33685760: reply ok 1448
> 20:00:51.969519 IP 192.168.0.79.nfs > netra.0: reply ok 1448
> 20:00:51.969618 IP 192.168.0.79.nfs > netra.0: reply ok 1448
> 20:00:51.969689 IP 192.168.0.79.nfs > netra.978: reply ERR 1448
> 20:00:51.969757 IP 192.168.0.79.nfs > netra.1345: reply ERR 1448
> 20:00:51.969828 IP 192.168.0.79.nfs > netra.0: reply ERR 952
> 20:00:51.969831 IP 192.168.0.79.nfs > netra.0: reply ERR 952
> 20:00:51.985618 IP netra.0 > 192.168.0.79.nfs: 0 proc-1888936968
> 20:00:52.793511 IP 192.168.0.79.nfs > netra.33685760: reply ok 1448
>
> I can mount the NFS shares/tftp from Linux OK (except for Linux's
> broken NFSv4 of course!)
>
> The bootparam/rarp/nfs/tftp services are all running according to svcs
>
> Can anyone help (or do you need more info?) as I have about a week to
> get this box co-located before my other web hosting contract runs out,
> so could do with getting on with securing the OS, not mucking around
> trying to install it!
>
> Cheers ;-)


Ok, Ill post this back here fyi, while im on lunch and since Im half
responsible for the netra, and Ive been trying to avoid having it
shipped to france to jump off my jumpstart server here like the
others.

Thanks for all the sugestions, we've tracked it down to a bug in the
vmware networking stack.
Theres a error in the dmesg on the solaris vm machine about how vmnet
uses one of the buffer structs incorrectly, and this could cause
problems on large IO operations as its not accessing the buffer
struct's in the correct manner, could be using the wrong offsets or
structure or anything. At the point of jumpstarting that it freezes,
the installer is we think calculating sizes of packages to check it
has enough space on the slices to install, and that probally
categorises as a "large io operation"...
I went along to the vmware forums and had a poke round, and the error
is a reoccuring one through various versions of vmware over the years,
and at one point the vmware tech on the forums says to "just ignore
the dmesg error, its not important.". Me and simon took a slightly
opposing view of the seriousness and impact of that as the nfs session
just seemed to peter out for no real reason.
So last night simon installed x86 solaris directly onto his laptop
(sacrificed to the cause!), and now the jumpstart server he installed
onto it is working flawlessly. Just got to configure it to do the SVM
partitioning tonight and itll be good to go get rehomed at the
weekend.

While it would be nice to just be able to have a vmdk on a dvd to have
a instant configured how we wanted it jumpstart server for emergency
rescue, I dont think the vmxnet stack is quite there yet...

Thanks again folks.
Phil P

From: Frank Cusack on 6 Mar 2007 14:35
On 6 Mar 2007 04:24:32 -0800 "MrFluffy" <phil(a)fluffycentral.com> wrote:
> Thanks for all the sugestions, we've tracked it down to a bug in the
> vmware networking stack.

Wow, that was a lot of work to find that out. Thanks for reporting
what the problem was.
-frank
From: Simon John on 7 Mar 2007 17:03
Well thanks for all the help folks, but I've fixed it now.

It seems there is a long standing VMWare bug that means that the
VMXNet driver falls over with large I/O under Solaris 10. VMWare keep
saying it's fixed, but it's still in Server 1.0.2

I installed Solaris on my laptop and the Jumpstart worked straight
away! I was quite impressed how well Solaris worked on the laptop
actually.

When I'm done tweaking the post-install scripts etc. I may tar up the
whole jumpstart structure and the /etc scripts and put it back on my
Linux box and see if I can get the Jumpstart to work from there too.
Gonna give up on VMWare though.

Next up is getting SAMP and Sendmail setup for my multiple domains,
then patching. I've got the 10_Recommended.zip patch cluster, but how
do I go about figuring out what further patches I need, like 120068-02
etc?

Cheers ;-)

From: Martin Paul on 8 Mar 2007 06:45
Simon John <simoninusa2001(a)yahoo.co.uk> wrote:
> Next up is getting SAMP and Sendmail setup for my multiple domains,
> then patching. I've got the 10_Recommended.zip patch cluster, but how
> do I go about figuring out what further patches I need, like 120068-02
> etc?

Sun provides a tool (smpatch) which has its quirks. For an alternative,
see my sig.

mp.
--
Systems Administrator | Institute of Scientific Computing | Univ. of Vienna
| http://www.par.univie.ac.at/solaris/pca/
Patch Check Advanced | Analyze, download and install patches for Sun Solaris
First  |  Prev  | 
Pages: 1 2 3 4 5 6
Prev: Cannot boot x4200 via dhcp (from another x4200)
Next: Solaris 10 tape drive permission issues...