Frequently Asked Question

SGI sells what they call the AWE6.5 which is the advanced workstation environment (Irix) version 6.5.x for around US$600.00 without a service contract, or around US$180.00 if you purchase a year's support contract on your system. Note that SGI may charge you to come out and inspect the system prior to starting a hardware support contract. This may or may not be such a good deal, as you may be able to pick up an entire used system for less than the price of a year's worth of SGI support.

There also seems to be a tremendous market for illegal copies of Irix on ebay. SGI sells directly on ebay as well, so SGI apparently looks the other way when it comes to the buying and selling of Irix media.

First go to http://techpubs.sgi.com/ and read the DISKS AND FILESYSTEMS book
to get an overview and some system specific information. Then you can follow
the steps below.

I would use a 1GB swap partition if you have a 4GB disk, 600MB is the
smallest you can get away with. If you only have a one or two GB disk,
buy a new disk. This part is important; 6.5 uses swap area as a staging
place for the installation. Too small a swap area will cause the
installation to complain. 

(These are my generic instructions) 

1. Put in the 6.5.x installation tools disk (See Notes at the end)
2. Get to the inst prompt. 
3. Go to the admin menu 
4. sh to get a shell 
5. unmount /root (if it's mounted) 
6. fx -x 
You will be prompted for the disk to fx:
7. > "device-name" = (dksc) <press enter>
8. > ctlr# = (0) <press enter>
9. > drive# = (1) <press enter>
10. > lun# = (0) <press enter>
You will get a menu at this point:
11. label/create/all 
12. label/sync 
13. repartition/rootdrive 
14. resize (Ideally you want 2GB of swap space for a 9GB+ disk) 
15. label/sync 
16. exi 
17. return 
18. exit 
19. Power cycle the system 

Then you install: 

1. The 6.5.x installation tools disk should already be in the CD-ROM drive. 
2.Boot the system by choosing install software 
3. While in inst, you run "From" to select /CDROM/dist and then run 'step'
to pick out the software that you want to install. 

If this is a new install, you will need at least the following disks: 

6.5.x Installation tools 
6.5.x overlays 2 of 2, 3 of 3 in 6.5.8 and above
(Note that the highest numbered Overlay disk (e.g. 3 of 3) you have to load
twice, once for /CDROM/dist and once for /CDROM/dist/unbundled). 
Applications 
6.5 foundation1 
6.5 foundation2 
6.5 Development Foundation 
6.5 Development Libraries 
ONC/NFS 
  
Put in any other disks you have, like compilers, libraries,
Performance Co-Pilot, etc. 

(Other versions have different number of disks, but you always put the
Installation tools in first, and you always put in ALL the disks that you
have. This may be 10 to 14.) 

(At some point in the installation, inst may not prompt you for more CDs.
Just use "Open" instead of "From" to add CDs to the list).

(For some disks, you may get a message that the software is incompatible with
the hardware. This is a bogus message, but you have to type /CDROM/dist/dist6.5
from the "From" or "Open" prompt to read the disk). 

4. Type conflicts to make sure that you have all the software needed for
what you want to install. 
5. Type go. 
6. Insert disks as prompted.
7. When finished type exit. 

The system will now boot. You then install anything else and configure
the system. 

After the system has booted and is in multi-user mode, it is a good idea to
run inst and put in the four 6.5.x disks (don't forget the unbundled
directory. Then type:
> keep *
> install upgrade

This will assure you that you have all the latest software loaded.

Notes:

If you have a new unused disk, then first you will need to partition it using 
standalone fx -x from the command monitor.

This command boots fx from an IRIX Installation Tools CD in a local
CD-ROM drive, where the CPU type of the system is IP19, IP20, or IP22 and
the CD-ROM drive is at drive address 4 on controller 0:

>> boot -f dksc(0,4,8)sashARCS dksc(0,4,7)stand/fx.ARCS --x

If you have an IP21, IP25, IP26, or IP27 then:

>> boot -f dksc(0,4,8)sash64 dksc(0,4,7)stand/fx.64 --x

. This command boots fx from an IRIX Installation Tools CD in a CD-ROM drive
mounted at /CDROM on a remote system named dist, where the CPU type of the
local system is IP21, IP25, IP26, or IP27:

>> boot -f bootp()dist:/CDROM/stand/fx.64 --x

See the techpubs URL listed above for more information.

I always do it manually, so I can tell what I've done. These are my
generic network instructions: 

Check the following. 

netstat -rn 
You should have a line that has default in it. If not, edit
/etc/config/static-route.options and put one in. 

/etc/resolv.conf 
Need either a domain line or a search line (not both), and one or more
nameserver lines. If you have less than 6.5.5 only use one nameserver line. 

/etc/nsswitch.conf 
Remove all NIS entries or comment out the lines with NIS only. 
Set the hosts line to files dns 
(This assumes that you are not running NIS) 

/etc/hosts 
Enter in lines for: The local hosts on your LAN, the gateway host, the
nameserver host(s) 

/etc/sys_id 
This should have the UNQUALIFIED host name. 

/etc/TIMEZONE 
Not really a network item, but it should be set correctly. 

/etc/config/netif.options 
This file can usually be left alone for a simple set up. Check to make sure
it's default You will see things like ${HOSTNAME} rather than specifics. 

/etc/config/ifconfig-1.options 
netmask 0xffffff00 broadcast xxx.xxx.xxx.255 
The netmask and broadcast address will be given to you by your ISP. 
The -1 must match the interface name
(rather than the actual interface: ef0 for instance). 
You may need to create this file. 

/etc/config/static-route.options 
Any static routes you may need, including the default gateway. 

/etc/config/ipaliases.options 
Put in any aliases that you want to use 

/etc/gateways Not used, remove it (unless you use routed or gated, but I
recommend you turn these off). 
/etc/init.d/network.local Not used, remove it if there is one 
/etc/init.d/network Used, leave strictly alone, make edits in
/etc/config/*.options files. 

Commands to run: 

chkconfig network on 

chkconfig routed off 

chkconfig gated off 

chkconfig autoconfig_ipaddress off 

chkconfig ipaliases on 

chkconfig nsd on 

chkconfig verbose on 

chkconfig savecore on 

chkconfig sar on 


For an non-graphical server (O200, O2K, Challenge): 

chkconfig desktop off 

chkconfig fontserver off 

chkconfig proclaim_relayagent off 

chkconfig proclaim_server off 

chkconfig  soundscheme off 

chkconfig webface off 

chkconfig windowsystem off 

chkconfig xdm off 

chkconfig videod off 

chkconfig visuallogin off 

All the above assumes that you have a fixed IP address. 
Routing additions may need to be made, depending upon your local LAN
configuration. 

After all this, reboot. (In this case easier than trying to figure out which
of the many init scripts to start.) 

Additional information can be found either in the help-->on-line books or
http://techpubs.sgi.com/

1. Log out of your session.

2. Log in as root (or get your sysadm to).

3. cd ~your_home_dir

4. ls -la |grep esk

4a. This will give you a listing like: 
drwxr-xr-x    8 jerry    user      4096 Nov 27 19:21 .desktop-hostname

5. mv .desktop-hostname junk (note the dot before desktop).

6.  Make sure of the following chkconfig settings:
desktop              on
mediad               on
noiconlogin          off
verbose              on
visuallogin          on
windowsystem         on
xdm                  on

7. Log out as root.

8. Log back in as yourself.

9. You should now have a stock desktop. If you don't, then you don't
have all the desktop subsystems installed.

10. If the root user is the problem, just log in as yourself, su -
and start at step 4.

1. su -

2. mkdir /usr/dist

3. ftp freeware.sgi.com
> cd dist
>i
>prompt
>hash
>mget *
>bye

4. find . -name "*tardist" -exec tar -xvf {} \;

5. inst
>from /usr/dist
>step
>conflicts
>go
>exit

When to use:

Perform these actions when the yellow service light comes on one of the
 JABOD chassis indicating a bad drive

Notes:

If these procedures are not followed, then you will loose all the data on the
JABOD and have to restore from tape! 

There are two methods: one which rebuilds the whole plex, and one which just
replaces a volume element. The first method is preferred, as only four
plexes are allowed per XLV volume and there are eight volume elements
on each plex, so you will run out of plexes shortly using the second method. 

Replacing a disk by rebuilding an entire plex:

   1. Determine the identity of the disk to be replaced by examining the
     chassis and matching the disk number with a device name. You 

   2. Log onto a host that the disk is attached to as root: su - 

   3. Verify the drive location (The "*" indicates the bad disk):
     # fccli getstatus -c 0 -e 0 -t 4

      Enclosure 0, status FAIL
       Vendor ID:     CLARiiON
       Product ID:    DISK ENC
       LCC SN#:       ms1034317700
       LCC Mfg. Date: 9824
       LCC code rev:  003
       FRUs:          10 disk slot(s)
                      1 fan slot(s)
                      2 power supply slot(s)
                      1 peer LCC slot(s)

       +---------------------------------------+------+
       | [ 0] FAIL                             |      |
       +---------------------------------------+      |
       |   |   |   | * |   |   |   |   |   |   | P    |
       | O | O | O | O | O | O | O | O | O | O | E  L |
       | K | K | K | K | K | K | K | K | K | K | E  L |
       | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | R  C |
       +---------------------------------------+      |
       |               FANS  OK                |  OK  |
       |      PS0  OK            PS1  OK       |      |
       +---------------------------------------+------+

   4. Determine the xlv object name for the jabod (/dev/xlv/???): df -k 

   5. Start xlv_mgr: xlv_mgr 

   6. Determine if this is the correct host: xlv_mgr> show -long object name.
     If not, log on to the other host, you must perform the operation from
     the host that the disk is active on (node=pain). 

   7. Create a script that you can cut and paste from to recreate the plex:
     xlv_mgr> script -write /var/tmp/scriptname object name. 

   8. Open up another window and load the file you created in the last step. 

   9. Determine the structure of the volume (In this example the volume name
     is "test"):
     xlv_mgr> show -long object test

     VOL test (complete)             (node=pain)
         VE test.data.0.0        [active]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d3s7 (8884447 blks)
         VE test.data.1.0        [empty]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d4s7 (8884447 blks)

  10. In this example drive 4 is bad; first detatch the plex:
     xlv_mgr> detach plex test.data.1 hold 
  11. Verify that the plex is detached: xlv_mgr> show -long all

     VOL test (complete) (node=pain)
         VE test.data.0.0        [active]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d3s7 (8884447 blks)

         PLEX hold
         VE hold.0       [stale]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d4s7 (8884447 blks)

  12. Delete the plex with the bad drive: xlv_mgr> delete object hold

     Object hold deleted

  13. Exit xlv_mgr: xlv_mgr> exit 

  14. Spin down the drive and remove:
     fccli bypass -c 0 -e 0 -t 4
     fccli remove -c 0 -e 0 -t 4
     Flags are -c = Controller, -e = Enclosure (chassis), -t = Tray (disk) 

  15. Replace drive and spin drive up: fccli insert -c 0 -e 0 -t 4 

  16. Check status of drive:fccli getstatus -c 0 -e 0 -t 4

      Enclosure 0, status FAIL
       Vendor ID:     CLARiiON
       Product ID:    DISK ENC
       LCC SN#:       ms1034317700
       LCC Mfg. Date: 9824
       LCC code rev:  003
       FRUs:          10 disk slot(s)
                      1 fan slot(s)
                      2 power supply slot(s)
                      1 peer LCC slot(s)

       +---------------------------------------+------+
       | [ 0] OK                               |      |
       +---------------------------------------+      |
       |   |   |   |   |   |   |   |   |   |   | P    |
       | O | O | O | O | O | O | O | O | O | O | E  L |
       | K | K | K | K | K | K | K | K | K | K | E  L |
       | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | R  C |
       +---------------------------------------+      |
       |               FANS  OK                |  OK  |
       |      PS0  OK            PS1  OK       |      |
       +---------------------------------------+------+

  17. Recreate the plex: xlv_make 

  18. Build plex ( be sure to use the original options from the script you
     created):

     xlv_make> plex new 

     new

  19. Create the volume element (repeat this step until all volume elements
     are created):

     xlv_make> ve /dev/dsk/dks0d4s7
     new.0

  20. End object creation: xlv_make> end

     Object specification completed

  21. Confirm Plex is created: xlv_make> show -long all

             Completed Objects
         (1)  VOL test (complete)                (node=pain)
         VE test.data.0.0        [active]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d3s7 (8884447 blks)

         (2)  PLEX new
         VE new.0        [empty]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d4s7 (8884447 blks)

  22. Exit xlv_make: xlv_make> exit

     Newly created objects will be written to disk.
     Is this what you want?(yes)
     Invoking xlv_assemble

  23. Reattach plex to original: xlv_mgr

  24. Verify the plex is there: xlv_mgr> show -long all

         VOL test (complete)             (node=pain)
         VE test.data.0.0        [active]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d3s7 (8884447 blks)

         PLEX new
         VE new.0        [empty]
             start=0 end=8884446, (cat)grp_size=1
         /dev/dsk/dks0d4s7 (8884447 blks)

  25. Reattach plex: xlv_mgr> attach plex new test.data

  26. Verify the plex is attached: xlv_mgr> show -long test

     VOL test (complete) (node=pain)
         VE test.data.0.0        [active]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d3s7 (8884447 blks)
         VE test.data.1.0        [empty]
             start=0, end=8884446, (cat)grp_size=1
             /dev/dsk/dks0d4s7 (8884447 blks)

         Vol: 1; Standalone Plex: 0; Standalone Ve: 0
          end=8884446, (cat)grp_size=1

             /dev/dsk/dks0d4s7 (8884447 blks)

         Vol: 1; Standalone Plex: 1; Standalone Ve: 0

  27. Exit xlv_mgr the revive begins on exit: xlv_mgr> exit 

fx -x
>label/create/all
>label/sync
>repartition/optiondrive
>label/sync
>exi

If your disk is larger than 9GB:

mkfs_xfs -l size=4000b /dev/rdsk/dks0d2s7

otherwise:

mkfs_xfs /dev/rdsk/dks0d2s7

The numbers are controller, disk, and slice.

The devices are normally made automatically, though you can
cd /dev
./MAKEDEV
to refresh them.

1. If it is an option disk:
        A. umount the disk.
        B. xfs_check /dev/dsk/dks#d#s7
                If xfs_check returns anything more than the next prompt
        C. xfs_repair
        D. Then run xfs_check again. If xfs_check is not clean repeat.

2. If it is a root disk:
        A. Power down the system.
        B. Insert the Installation Tools disk.
        C. Power up, enter maintenance mode, and choose install system software.
        D. When you get to the inst prompt (ignore messages about additional disks).
                > keep *
                > admin
                > umount /dev/dsk/dks0d1s7
                > sh
                # xfs_check /dev/dsk/dks0d1s7 and repeat 1B, 1C, and 1D
                # exit
                > return
                > quit
        The system should boot at this time.
        Once booted, it would be a good idea to run inst and do a keep *
        install same (You will need to put in ALL your disks at this
        point--14 or so) to check that all the files are intact, and fix those
        that aren't.

1. Boot the system.
2. Stop for maintenance.
3. Choose install system software
4. Put in the CD (marked Installation Tools)
5. Wait for the inst prompt
6. keep * (not required but it will eliminate messages about uninstalled software)
7. admin
8. sh
9. cd /root/etc
10. ed passwd  or ed shadow 
11. Using the ed commands, remove the passwd hash.
12. cd /root/etc/default
13. ed login
14. Using the ed commands, make sure that passwords are not required.  
                  (change some YESs to NOs)
15. exit
16. return
17. quit
18. reboot

You will now have a system with no root password. Be sure to reset the items once you are 
in as root.

1. Raw swap partition--this is the fastest, but requires that you repartition the disk
to change it. 

2. Preallocated file--this is the second fastest, you use mkfile to create a file of
some length and then point swap to it. It has the advantage that you can create it at
any time, and it is all in one contiguous set of blocks on the disk. It still takes up
disk space, and is not accessable from the miniroot for installation purposes. 

3. Zero length file--You use touch to create a file and then point swap to it. It has
the advantage of not using space until you need it. It has the disadvantage of possibly
filling up the filesystem unexpectedly, and having extents. 

4. Virtual swap--This isn't really swap space at all, you are just telling IRIX that
you have this much space when you don't. It's generally good for applications that
spawn a bunch of processes rapidly which would allocate a lot more swap than you
actually have (mail servers would be a good example). Some applications will die horribly
if this is turned on (Photoshop for instance). 

1. You want to upgrade only what is already installed:

        su -
        inst
                Open up all four disks and /CDROM/dist/unbundled
                Choose done
                keep *
                install upgrade
                conflicts
                go
                exit
        versions changed
        Edit the files shown by versions changed
        reboot

        versions changed        versions changed
        Edit the files shown by versions changed
        reboot

2. You want to upgrade and add new products
        su -
        inst
                Open up all for disks and /CDROM/dist/unbundled
                Open up all the other base, compiler, and etc. disks, but not
                        6.5 base installation disk, and not any 6.5.x disk except
                        the release you are installing
                Choose done
                keep *
                install upgrade
                step
                        At this point, go through and pick all the new software
                        you want
                conflicts
                go
                exit
        versions changed
        Edit the files shown by versions changed
        reboot

I use my home Octane mostly for development and web stuff along with a little
DTP. Typically, I have over 200 processes running before anyone logs in. I
have the CADduo system which gives two keyboards, mice and monitors, so that
two people can use the system at the same time, just like they had their own
workstation.  Subjective performance is that I can't tell when the other
person is on.

I'm running three types of weberservers, two proxy server, ldap server, SQL
database, calendar server, mail, named, plus regular applications like
netscape and sendmail. The Octane just eats it up. If I were to put this on
a PC or Mac platform, I would need a minimum of ten systems the total would
come to far more than I paid for the Octane--I would also have to purchase
more software and configure the systems to all work.

Let's not talk about how I would have to spend my weekends admining the
systems (with ten systems, you always have at least one system that has
problems at any given time), or the electricity it would use (Intel chips use
about ten times the electricty that MIPS chips do), or the additional network
hardware I would need, and since I'm in Texas the a/c bill would rival the
space program.

Then to top it off, I would have to purchase the entire system again in two
years. So, each of the ten system would likely cost 8 to 15K (you have to
buy drives and software. One of the big places that people get sucked in is
that they see a $2000 price for a PC and assume that's all they pay, for a
high quality one, you will pay much more, then add software, extra cards,
good monitor, ...), but even if I could get by for $4K, that's still 40K,
about the price for a fully equipped new Octane. So going over the figures
for eight years of computing (not including software): 

Octane $50K (Everything you could want--assuming you've done your homework
and gotten a good deal). 

Year 3, 5, 7: Upgrade processors: $5K*3 upgrade = $15K 

Maintenance contract $1500*7 = $10.5K  (does not include compiler).

Total cost of ownership: $75.5K 

Estimated trade in value for Octane at the end of eight years: $10K 
Adjusted cost of ownership 65.5K 

(Note that my previous computer was an Indy which I kept from early 94 to 99,
I got back 80% of the original base purchase price as a trade in for the
Octane, try this with a PC). 

10 PCs * $4K = $40K * 3 upgrades: $120K 
Maintenance from a third party ( $400 * 10 * 5 ) = $20K--The five comes from
the five years that aren't covered by warranty.

Total cost of ownership $140K.

xlv_make

> vol myvol
> data
> ve dks1d2s7
> end
> create
> exit

xlv_assemble
mkfs_xfs -l size=4000b /dev/rxlv/myvol
 
To stripe an XLV volume from multiple disks:

xlv_make

> vol myvol
> data
> ve -stripe -stripe_unit 128 dks1d2s7 dks1d3s7 dks2d2s7 dks2d3s7
> end
> create
> exit

xlv_assemble
mkfs_xfs -l size=4000b /dev/rxlv/myvol