Re: device packages?!?

Lots of information, but work through stuff one step at a time.

Don't get overwhelmed. Digest in bite-sized chunks. :)

If any questions, just ask.

In article <1137627315.891512.19090@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>, cc
<christiancolborn@xxxxxxxxx> wrote:
>
> i'll be back on here in a couple hours, but dan, if youre gone by
> then... what do you think is the best way to get the drivers for
> those NIC's & fibre-channel's installed? i have to get those things
> working one way or another tonight...(is there a way to test theyre
> working, besides 'diag' without connectivity?)

These filesystems are full; we need a little extra room, so grow each
filesystem listed by 10 512-byte blocks (aka 5K total per fs):

# chfs -a size=+10 /tmp
# chfs -a size=+10 /usr
# chfs -a size=+10 /var

(chfs is the AIX equivalent of Solaris growfs.)

We are clearing out *just* enough room so that rest of the commands
won't fail when they create temporary files.

Also, I'd suggest removing any junk from all three filesystems. (rm
stuff that you can safely remove. If you're not sure, ask in a separate
thread.)

Now we double check to make sure we have any existing data moved out of
the way:

# ls -l /usr/sys/inst.images

If above ls command returns no files, great. If it has any files in it,
then do this:

# mv /usr/sys/inst.images /usr/sys/inst.images.old

Then we check something important:

# lsvg rootvg | grep "PP SIZE"

If PP size is set to 64 MB, then do this:

# mklv -c2 -y lpplv rootvg 10

That will create a 640 MB logical volume (LV) for you.

-c2 specifies two copies (for a mirrored setup).

-y lpplv specifies name of LV

rootvg specifies which VG to carve the LV out of.

10 specifies 10 PPs to take for making the LV.

If PP size is set to 128 MB, then do this instead:

# mklv -c2 -y lpplv rootvg 5

5 x 128 MB = 640 MB

You get the idea. For any PP size, multiply it by whatever number to get
640 MB. I'm not sure you need that much, but doesn't hurt, especially
since you have a *LOT* of unused disk space right now.

And now, to create and mount the filesystem itself:

# crfs -v jfs -d /dev/lpplv -m /usr/sys/inst.images
# mount /usr/sys/inst.images

If we had any files in /usr/sys/inst.images.old then move them to the fs:

# mv /usr/sys/inst.images.old/* /usr/sys/inst.images/

Do you already have a cdrom entry? Check:

# grep -p /cdrom /etc/filesystems

If that returns nothing, then let's make a CD entry so we can mount it.

# mkdir /cdrom
# crfs -v cdrfs -d /dev/cd0 -m /cdrom -p ro

(This will add a /cdrom stanza in /etc/filesystems, the AIX equivalent
of /etc/fstab.

cdrfs is also the AIX equivalent of Solaris pcfs filesystem.)

Now mount the CD:

# mount /cdrom

Which directory has the files on CD? Check.

# ls -alR /cdrom | more

Whereever the package files are, copy them to /usr/sys/inst.images.

Sometimes you can tell if they are package files because they have a
..bff filename extension (BFF = Backup File Format).

Or sometimes you can do 'file *' in a particular directory and it says
'backup/restore format file'.

(You normally do not need to copy from a CD to a hard drive; AIX can
read packages without having to mount a CD, but in your situation, the
CD image you have is missing the TOC and possibly other things... so we
will make sure everything looks ok this one time.)

Then build the Table of Contents (TOC) file:

# inutoc /usr/sys/inst.images

*NOW* you can install stuff from there.

You can use either 'installp', 'smitty install', or 'cfgmgr'.

Also, see if you're missing any of these packages by doing this:

# lslpp -L | grep devices.pci.df1000f9
# lslpp -L | grep devices.pci.df1000f7
# lslpp -L | grep devices.pci.df1080f9
# lslpp -L | grep devices.common.IBM.fc
# lslpp -L | grep devices.fcp.disk

(lslpp is the AIX equivalent of Solaris pkginfo.)

Let us know if above lslpp had anything for above five filesets? And if
yes, what version (5.2.0.something).

If necessary, install missing stuff with:

# installp -aXg devices.pci.df1000f9 devices.pci.df1000f7 \
devices.common.IBM.fc devices.fcp.disk devices.pci.df1080f9

Leave out any names of filesets that you already have.

(installp is the AIX equivalent of Solaris pkgadd.)

Check the lslpp -L again for anything you installed. See it appear now?
If yes, what version?

Retry the configuration manager probe:

# cfgmgr

(cfgmgr is the AIX equivalent of Solaris devfsadm.)

Afterwards, you can check to see if AIX saw the device:

# lsdev -C | grep fc

You should have devices like 'fcs0' appear (FC adapter #0). They should
also be in the Available (ready to be used) state.

Defined means "It's there but I'm not sure if it can be used unless you
change fix or enable it".

Also, to check what network interfaces you have installed:

# lsdev -Cc if

Ignore any et lines (et0, et1, etc) or lo0. You want to see what 'en'
lines appear -- en0, en1, etc.

Then make sure you're looking at the right one by doing:

# entstat -d ent1 | head -5

Make sure the 'Device Type' looks reasonable. You're just making sure
this isn't a built-in device or that you're looking at the wrong
interface.

It could be named ent1, could be ent2, ent3, or who knows what. That's
why you're triple checking with entstat to see which one is it --
assuming the device driver installed ok.

Other people might have an opinion on how to do quick testing without FC
or ethernet connectivity... but my guess is that if you get the devices
to show up at all, they should work at the customer site.

Also, will this be used on a gigabit ethernet network or 100 mbit LAN?

If 100 mbit LAN, find out if the network port will be autonegotiation or
hardcoded on the switch side.

If hardcoded -- let's say, for 100/full, then on the AIX side, do this:

# chdev -l entX -a media_speed=100_Full_Duplex

(where entX is the interface -- ent99, ent3, whatever)

If it's autonegotiation on switch port, do not hardcode the network
interface's speed and duplex.

If it's gigabit ethernet, leave as autonegotiation because that is what
the gig-e standards requires.

Also, once you get the FC device drivers loaded, check the firmware
revision on the FC card(s):

# lscfg -vl fcs0

Look at the ZB field. Ignore first three characters of the value. If
it's a Feature Code 6239 (aka FC6239) card, it needs to be at least
version 1.81X1.

I *think* the '5704' you mentioned earlier maps to a FC 6239 card.

Go here:

http://publib.boulder.ibm.com/infocenter/pseries/v5r3/topic/com.ibm.pseries.doc/hardware_docs/selectbygeneralserviceoptions.htm

Download the 'Adapters, Devices, and Cables for Multiple Bus Systems'
PDF file from there.

Ignore the Microchannel (MCA) stuff 'cause you probably don't have
MCA-based systems, and the pSeries are all PCI based, anyway.

Look at page 125 of that PDF file. It says FC 6239 is a single port 2
Gb/sec FC adapter, and mentions 'Type 5704'.

On page 129, it says FC 5706 is a dual-port 10/100/1000 Mbit Ethernet
adapter, and also type 5706. Sounds right to me, too.

Also go to that link I gave and download the 'PCI Adapter Placement
Reference' PDF file. DON'T get the Linux one (first). Get the second PDF
file!

Open that file and click on the entry for the pSeries 630 (6C4/6E4).

Look at the table. It says that the FC 6239 (aka type 5704; that's the
FibreChannel card) should be put in slot 4 first if this is a 4-slot
p630.

If slot 4 is full, put it in slot 3. If 3 is full, put in 2. If 2 is
full, put in 1. If 1 is full, buy a bigger machine. :-)

If this is a 6-slot p630, the preferred order is 3, 4, 5, 6, 1, 2.

You can look up the entry for the FC 5706 ethernet card you installed.
The table says same order as for the FC 6239 card, too.

That table basically tells you how to optimize your machine for
performance by putting cards on their own high speed busses instead of
fighting each other for internal bus bandwidth and stuff like that.

-Dan
.

Relevant Pages

  • Re: Cannot boot using Cdrom for Install--Pse Help!!
    ... >>Do you have another box on which you can mount the cd to find out if it ... > Unfortunately, I don't have another Sun System, but have a Red Hat 9.0 ... I am getting thesse new errors when tried to install using a new ... In what hardware are you trying to install Solaris? ...
    (comp.unix.admin)
  • Re: NFS problems with Linux mounting a file
    ... but flash is not limited to the core Solaris product. ... > have one Linux box working here so I can't test mounting a 3GB file Linux ... No, from what I tried out, Linux is unable to mount files at all. ... although you might need to install a larger HDD and more memory. ...
    (comp.os.linux.networking)
  • Re: NFS problems with Linux mounting a file
    ... but flash is not limited to the core Solaris product. ... > have one Linux box working here so I can't test mounting a 3GB file Linux ... No, from what I tried out, Linux is unable to mount files at all. ... although you might need to install a larger HDD and more memory. ...
    (comp.sys.sun.admin)
  • unable to install from CD: failure to mount once kernel installed
    ... cleanest possible install. ... ran netinstall off floppies accepting all defaults: still unable to mount ... booted the Linux box from the hard disk into the testing Debian distro, ... allow me to turn off shadowing for either the video BIOS or the main BIOS. ...
    (Debian-User)
  • Re: Resizing Partitions, Losing Windows XP...
    ... install applications ... Could I do a dump of my current FreeBSD ... to mount the partitions read only. ... now to restore there are better ways to do it but this ...
    (freebsd-questions)