Introduction:
This paper contains information regarding the initial opening, setup
and deployment of a Sun Server SunFire v100. The paper may or will
contain technical terminology including detailed information
regarding every step taken on how everything was performed.
Opening:
The server came wrapped in bubble wrap, packing peanuts, air bags and
was inside a cardboard box, the item is as described and as shown in
the pictures provided by the eBay seller, it is a SunFire v100
Server, no rails included, has mounting ears, one of them has some
oxide, on the back there is a Sun card which is used for
configurations, the inside is in great shape and clean, seems as if
it had been recently pulled out of production. Inside there are two
IDE hard drives, 40 GB each, with 2GB of PC133 RAM split up in 4
modules of 512MB each one. The processor seems to be mounted with the
heatsink but to avoid moving or messing with anything at this point,
no cable has been included or any other accessories are available
with this item. Final cost of item was 35 USD on July 2018 of which
12.50 USD was the cost of the item itself, the rest was for shipping
and handling. This item was opened on July 14th, Saturday,
2018.
Setup:
Connecting the server to a light source did not turn it on as it is
with most servers nowadays, there is a switch but it was not flipped,
to connect to the console a Toshiba Satellite machine with Windows 98
and serial port was used, a DB9 to RJ45 cable was used to connect to
A/LOM on the back of the server, as it turns out LOM indicates that
the server is on stand by mode through most of the commands and can
power on through this method, doing so does power on the server and a
full boot up took the server onto Solaris 5.10 which was configured
with an AT&T business IP address, there was no information
available or anything critical leftover on the server two hard
drives.
Rebooting the server did not bring the system back to its original
mode, it would seem as if the server has fallen back into net mode
for no apparent reason and wants to forcibly boot into a PXE Server
or some other server that will provide it a kernel so that it can
work (this seems to be normal from back in the day on what Sun was
doing). Going through the options it seemed logical to install
another OS into the Server since Solaris was not working properly and
the version, while still supported, was of no use to me and I would
rather have anything else than this, such as Debian or Gentoo.
One of the initial issues found with the system, aside from the
problem of not booting up, was making the Sunfire v100 stop trying to
boot through net and go into anything else, on the LOM prompt this
can be done first typing on the LOM prompt bootmode forth, this will
stop the current boot process and take us to the OpenBoot prompt, in
order to get to this prompt the Sunfire v100 has to be powered on
(LOM>poweron).
Once the server has been halted it is easy to boot from the CD (or
DVD, depending on what you may have on your machine, by the looks of
it this has been somewhat rigged to use a slim CD-ROM but it could
easily use a CD/DVD Burner from the looks of it, while I do have some
slim readers/burners I have no intention of burning any CD/DVDs from
this machine at this point), type the command boot cdrom and it
should go into the CD, in my case it would not read the CD or would
give me the error Fast Data Access MMU miss with, by all means, could
mean that anything is not working properly inside the server and thus
this is why it is not booting up.
Following some instructions I first disabled the auto-booting, this
can be done (as indicated in this link:
https://it.toolbox.com/question/fast-data-access-mmu-miss-in-sunfire-v120-061010)
by doing setenv diag-level? max, then, setenv auto-boot? False, then,
reset-all. If all commands are followed correctly the server will
throw up all the things that it is currently doing and will not boot
towards anything, it will simply reach the OpenBoot prompt and stay
there (I have yet to find a way for ALOM to forward logs to a SIEM or
any other log storage method so that these can be read elsewhere).
While doing so I got a warning about a possible issue with RAM, DIMM0
was the probable culprit per the initial boot-up so after shutting it
down I removed it and booted back, the error still came up but I
ignored it, at this point I wish I had not lost a whole day figuring
out what to do when the error was in front of me the whole time.
Booting to a CD worked, I had issues probing for the drives but I
decided to ignore it (later this issue would be resolved through a
different method as provided by Oracle on some documentation they
still have on their website). It took about a whole day for Debian to
run the installation and in the end crashed, OpenBSD and NetBSD did
the same, I had looked at a YouTube channel, Backyard Tech, in which
he explains to another user the possible issues he could be facing
with a Sunfire v120 (originally I had only intended to buy the
Sunfire v120 but I found the Sunfire v100 at about the same price so
I also went for it, the Sunfire v120 is another story that required a
whole different document to explain the ordeal it will be to get it
to work). One of the things he had mentioned was to burn the CD as
slow as possible to have the least amount of errors, one of the steps
I took to make sure that the system would load, but had mixed
results. Having the Sunfire v120 at hand I changed the CD-ROM but
found the same results with the CD which made me suspicious of the
internal controller.
Since I had no way to change the controller or do anything else at
this point I went back to square one and removed all the RAM, in the
end the problem with the server was indeed the RAM, it would seem
that either some of the sticks or ports were not working properly,
leaving only one stick of RAM, on DIMM 3 (remember that RAM is
inserted from back to front not the other way around, this is
important and is notated on the documentation and the images provided
on the cover plate of the Sunfire v100). At the moment 512.
A link with assistance on configuration of OpenBoot can be found
here, the website contains all the necessary information to properly
configure OpenBoot to boot to the correct disk once formatting has
been performed and also to troubleshoot properly since I was
originally not able to probe for SCSI or IDE drives, once the
procedure was followed everything was showing up and working fine,
boot up now takes me directly to Debian when turning it on so there
is no real necessity at this point to go through the Windows 98
machine to work on LOM, this is being used only to monitor resources
consumption at the moment:
https://docs.oracle.com/cd/E23824_01/html/821-2731/gkkvd.html.
With Debian installed there is another issue, installing it was no
issue, interestingly it would not pickup the repository for sparc64
from the ports repository, could not add it manually either at that
point so running with whatever the netinst CD had was what needed to
be done at this point, the site for the sparc64 port indicates that
some commands need to be run in order to get the proper software
installed once the source.list file is corrected (all the information
can be found here: https://wiki.debian.org/Sparc64).
The problem in the end is that the security repository for sid is not
working, not sure why this is, and there is a considerable lack of
packages available for the platform. While I could install apache2
and the compilers through build-essential, python3 and pip3, the
packages for MariaDB Server 10.1, MongoDB or any other database aside
from PostgreSQL are missing, as a matter of fact most of the
libraries required to build MariaDB are missing from the Debian
repository, for the last two days (Thursday 19th and
Friday 20th) I have been compiling libraries to get
MariaDB to run on the Sunfire v100. I know I could use SQLite or
PostgreSQL but I would like to have MySQL/MariaDB available in my
current Debian build
No comments:
Post a Comment