[mamorim01]

Greetings everyone,

I have been reading the forums for some time, I meant to register earlier on shortly after nekochan's demise, but things kept getting in the way. I own a few SGI units which I have been gathering over the years, and I am slowly bringing them back to service. Most of them were non functional at the time I procured them, missing or dead hard drives, some of them had minor issues of some kind such as the Indigo2 that crossed Europe inside a cardboard box with no padding (sad face, plastic bits all over the place). Two units are practically restored (an Octane and the Indigo2) and a third one is getting there (an old and humble Indy). If anyone is interested, I would be happy to post pictures and hinv outputs.

I do have a quick couple of questions which I hope someone can possibly share some light into.

- I have been having some issues with the hard drive in the Octane, which was shipped with the unit with a clean 6.5 install. I managed to reset the root password through miniroot (booting from DINA) and the system is practically functional although there is a bit of trashing and non-responsiveness at times. Running xfs_check shows a bunch of errors which I would like to repair, if possible. Running xfs_repair requires the drive to be unmounted so, seeing as I do not have access to a CD-ROM, my only chance seems to be running xfs_repair through a miniroot from DINA. This does not appear to work, as I seem to be unable to unmount anything from miniroot (ie, /dev/root: device busy or /dev/dsk/dks0d....: device busy). I get the exact same errors as if I had booted from the main drive instead. Short of pulling the drive from the unit and mounting it in other computer as second drive, is there an easy way to run xfs_repair in the Octane so as to fix the hard drive?

- As for the other question, the Indigo2 (teal color) was originally fitted with a GR2-XZ graphics card, and it is running 6.5.22. I eventually managed to find an Elan card for a very reasonable price and went on to install it in parallel with the XZ hoping to be able to run a dual head configuration at some stage (I moved the XZ card to the originally empty slot, and installed the Elan card in the slot which was originally occupied by the XZ card). Would this kind of configuration possibly work? The system detects both graphics cards but I am unable to get any video out of one of the cards. I have not attempted a reinstall of the OS, in case it might matter, which I think it could.

Thank you all for reading this and for carrying the torch, it is great to have a repository of information for these machines. I keep a close eye on your efforts to port more software and bring development tools up to date. I sure hope I could do something along those lines, but my expertise is very far from having meaningful coding skills. Hope I can eventually provide something in return nonetheless.

Cheers!

[epitaxial]

The drive in your Octane is likely failing. I would replace it.

[dexter1]

The Indigo2 riser board has three GIO64 connectors, of which the top two connectors are GIO1 and the bottom connector GIO0.
Putting the Elan in GIO0 and the XZ in GIO1 should in principle work: Go to the PROM and check 'hinv' , if both are detected, you're fine.

Both cards are Express graphics, so the IRIX OS should not need additional drivers or replacing the X.eoe subsystem. But you probably do need to reconfigure the kernel:

1)  ioconfig -f /hw
2)  autoconfig -f
3)  reboot system

if that doesn't help, probably a reinstall should be in order.

Jan-Jaap's post below raises an important point. Your PSU should be capable of running dual-head. What is the part number of the PSU?

[jan-jaap]

Hi there!

- I have been having some issues with the hard drive in the Octane, which was shipped with the unit with a clean 6.5 install. I managed to reset the root password through miniroot (booting from DINA) and the system is practically functional although there is a bit of trashing and non-responsiveness at times. Running xfs_check shows a bunch of errors which I would like to repair, if possible. Running xfs_repair requires the drive to be unmounted so, seeing as I do not have access to a CD-ROM, my only chance seems to be running xfs_repair through a miniroot from DINA. This does not appear to work, as I seem to be unable to unmount anything from miniroot (ie, /dev/root: device busy or /dev/dsk/dks0d....: device busy). I get the exact same errors as if I had booted from the main drive instead. Short of pulling the drive from the unit and mounting it in other computer as second drive, is there an easy way to run xfs_repair in the Octane so as to fix the hard drive?

FWIW, I don't think there's xfs_repair in the miniroot. What you could do is boot into 'fx' from the PROM, and then run a (read-only!) sequential exercise of the disk: '(exe)rcise > (seq)ential > (ro) > select whole disk > go'. This would force it to map out bad blocks. Then run xfs_repair, using another system. But if that disk has bad blocks, I'd still replace it. You could try to cone the disk to a new disk if there's something you wish to preserve.

- As for the other question, the Indigo2 (teal color) was originally fitted with a GR2-XZ graphics card, and it is running 6.5.22. I eventually managed to find an Elan card for a very reasonable price and went on to install it in parallel with the XZ hoping to be able to run a dual head configuration at some stage (I moved the XZ card to the originally empty slot, and installed the Elan card in the slot which was originally occupied by the XZ card). Would this kind of configuration possibly work? The system detects both graphics cards but I am unable to get any video out of one of the cards. I have not attempted a reinstall of the OS, in case it might matter, which I think it could.

Beware that half a dozen different types of power supplies used to exist for the Indigo2. Dual head requires a dual-head capable PSU to supply the extra current. Also, even though there are three or four GIO connectors, logically there are only two slots, and no two cards can share the same slot so placements of the cards is important. See: http://archive.irix.cc/sgistuff/hardware...ml#options

You cannot mix different generations of cards (e.g. Express and IMPACT) but you're not doing that, so my guess is that the placement of the cards is not correct.

[mamorim01]

Thank you all for sharing your thoughts. I do agree that the safe way to move forward is to simply replace the drive. I was hoping for a way to schedule xfs_check/xfs_repair at boot similar to how fsck can be scheduled in linux, as a way to hold onto the drive for a while longer. Besides a few things I would like to preserve, I seem to be having issues with the units loosing connection with DINA after booting sash/fx, anything past that point and they do not seem to see the boot server, at all, so I am for the moment unable to carry out a fresh install from the network. Moreover, the sled that contains the drive seems to be stuck inside the Octane tray, and I am a bit wary about the possibility of the plastic caddy breaking off. I will investigate both issues further and will install a new drive when I have sorted both things out.

As for the dual head configuration, I was aware there could be issues with the standard PSU, and that the SolidImpact cards are specially demanding on that regard. While no power ratings are visible, I am working with the assumption that the PSU in question is a #6064470 by ITT PowerSystems which, if I am not mistaken, seems to be rated at around 1 kW (around 500 W on the DC rails). The two warning labels match with pictures I found online, and the distribution of screws on the top of the case would be exactly the same. The power rating label on this PSU is affixed to one side of the case which is in contact with the computer case, hence concealed unless you physically pull the PSU, but I am fairly certain it is the same model.

I tried reconfiguring the kernel and it seems to have done the trick. Either that triggered an update, or something else changed since the last time I tried which was, admitedly, some time ago. I now get video from GIO0 (the one at the bottom) at PROM and some sort of out of sync signal from GIO1 (the one on top). However, the moment IRIX boots I get two good signals on both monitors, and I can access both desktops with no issues whatsoever.



I will check the PSU in more detail to confirm it will be safe to run this configuration. I have unfortunately not found much information online regarding the different models, their power ratings or power consumption estimates for different configurations.

As I said, thanks much, this was good info all around!

[Trippynet]

If you can get another drive, you could try to clone it with xfsdump or similar? I usually have 3 different drives I rotate through my systems from time to time in order to do a dump of the system drive so I've got a few ready-to-go backups on-hand. Then again, due to decommissioning lots of old servers, I've got a big pile of spare drives

[mamorim01]

I will definitely try to do that. Hopefully the existing errors will not prevent the cloning from happening. Need to get the drive out of the sled first...

[Infiltratr]

These are the instructions I use for drive cloning in my octanes.

I use the dd section because its simple and it works for me.

http://www.cspry.co.uk/computing/Indy_ad...ackup.html

[mamorim01]

Cool, I will keep it at hand, thanks!

[gijoe77]

These are the instructions I use for drive cloning in my octanes.

I use the dd section because its simple and it works for me.

http://www.cspry.co.uk/computing/Indy_ad...ackup.html

I am under the understanding that using dd is more dangerous because it doesn't deal with bad sectors/blocks in a robust way, and you may have data corruption issues or even an unbootable system depending on what is corrupted.  This was always my understanding and why it was preferable to use xfsdump/xfsretore.

There was a disk clone script on Nekochan that I never got a chance to save (or at least now I Can't find it - can someone post it please?)

Here is a script I found a long time ago somewhere, it works fine

Code:

#!/bin/sh
# program to copy root & usr partitions and create a new bootable disk
# Note that this script is limited to SCSI drives, and only those
# on controller 2.
# It makes the root & usr filesystem on partition 0 & 6 for destination drive
# and install sash, ide & fx in the volume header of  the destination drive.

if [ $# -eq 0 ]
then
        echo "Usage $0 [disk_number]"
        exit 1
elif [ $1 -eq 1 ]
then
        echo "The disk number needs to be greater than 1."
        exit 1
elif [ $1 -le 1 ]
then
        echo "The disk number needs to be greater than 1."
        exit 1
else
        NUM=$1
fi

echo "\nSource SCSI (system)disk 1"
echo "\nDestination SCSI controller:  \c"
read SCSI_C
echo "\nDestination SCSI Controller channel "$SCSI_C" disk "$NUM""
echo "\007OK? (n)\c"
        read answer
        case "$answer" {
        y|Y|yes|Yes|YES) ;;
        *) echo "\007\n\nAbort"; exit 1;;
        }


echo "\nChecking SCSI disk"$NUM".."
hinv -c disk | grep -s "unit $NUM on SCSI controller $SCSI_C" > /dev/null 2>&1
if [ $? -ne 0 ]
then echo SCSI disk $NUM not found on controller $SCSI_C
        exit 1
fi

(prtvtoc -s -h /dev/rdsk/dks"$SCSI_C"d"$NUM"vol | grep "0.*xfs" ) > /dev/null 2>&1
if [ $? != 0 ]
then
        echo "reformat disk $NUM on SCSI controller $SCSI_C\n"
echo "for a system disk (no partition 0 found)"
        exit 1
fi

/etc/fsstat /dev/dsk/dks"$SCSI_C"d"$NUM"s7 > /dev/null 2>&1
s7=$?
/etc/fsstat /dev/dsk/dks"$SCSI_C"d"$NUM"s6 > /dev/null 2>&1
s6=$?
/etc/fsstat /dev/dsk/dks"$SCSI_C"d"$NUM"s0 > /dev/null 2>&1
s0=$?
if [ $s7 -ne 3 -o $s6 -ne 3 -o $s0 -ne 3 ]
then
        echo "\007SCSI controller $SCSI_C disk $NUM appears to\n"
echo "have a valid filesystem, overwrite it? (n)\c"
        read answer
        case "$answer" {
        y|Y|yes|Yes|YES) ;;
        *) echo Disk $NUM not added; exit 1;;
        }
fi
#-------------------------------------------------

echo "\nStart disk clone..."
umount /d
mkdir /d
mkfs -d name=/dev/rdsk/dks"$SCSI_C"d"$NUM"s0 -l internal,size=2048b -b size=1k
labelit /dev/rdsk/dks"$SCSI_C"d"$NUM"s0 / sgi
mount /dev/dsk/dks"$SCSI_C"d"$NUM"s0 /d
df -k
echo "\ncopy root partition..."
cd /; xfsdump -l 0 -J - / | (cd /d; xfsrestore - .)

fsstat /dev/usr 1> /dev/null 2>&1
stat=$?
if [ $stat = 2 ]
then
        mkdir /d/usr
        mkfs -d name=/dev/rdsk/dks"$SCSI_C"d"$NUM"s6 -l internal,size=2048b -b
size=1k
        labelit /dev/rdsk/dks"$SCSI_C"d"$NUM"s6 /usr sgi
        mount /dev/dsk/dks"$SCSI_C"d"$NUM"s6 /d/usr
        df -k
        echo "\ncopy usr partition..."
        cd /usr; xfsdump -l 0 -J - /usr | (cd /d/usr; xfsrestore - .)
fi

echo "\ninstalling sash, ide & fx in volume header....."
if [ ! -r /stand/sash ]; then /etc/dvhtool -v get sash /stand/sash /dev/rdsk/dks0d1vh ;fi
if [ ! -r /stand/fx   ]; then /etc/dvhtool -v get fx   /stand/fx   /dev/rdsk/dks0d1vh ;fi
if [ ! -r /stand/ide  ]; then /etc/dvhtool -v get ide  /stand/ide  /dev/rdsk/dks0d1vh ;fi

/etc/dvhtool -v c /stand/sash sash /dev/rdsk/dks"$SCSI_C"d"$NUM"vh
/etc/dvhtool -v c /stand/ide ide /dev/rdsk/dks"$SCSI_C"d"$NUM"vh
/etc/dvhtool -v c /stand/fx fx /dev/rdsk/dks"$SCSI_C"d"$NUM"vh
echo "\nNew disk number SCSI:"$SCSI_C" disk:"$NUM" ready to use."

echo "\nDONE....."
exit 0