Extending LUNS ? Part II

[Simon M Appleby]

For Tom Geudens :

Tom,

You stated in your reply that you like to see the posted "VGDISPLAY" for the offending volume groups. I have managed to obtain a printout and thusly I have attached it for you to looke over.

I have question though. How can I work out from this print out how much space (Mb) I have remaining to this Raid device. Once I can determine the space remaining then I can, I suppose destroy and recreate the LUN, VG etc, with the remianing space.

Can you show me the formula (If any) on how this calculation works.

Thanks once again.

To All whom replied I shall allocate points when I have completed this request.

Thanks to all once again. :-)

[Scott Van Kalken · Accepted solution]

The things you need to look for are the PE stuff.


VG Name /dev/vg00
VG Write Access read/write
VG Status available
Max LV 255
Cur LV 10
Open LV 10
Max PV 16
Cur PV 2
Act PV 2
Max PE per PV 5120
VGDA 4
PE Size (Mbytes) 4
Total PE 2046
Alloc PE 1618
Free PE 428

PE size - specified in megabytes. This is the size of each Physical Extent on the disk. Default if 4mb

Total PE - this is total number of 4mb extents you have on the disk.

Allocated and Free are important too.

If you take the Free PE X PE Size we get 428 X 4 = 1712mb

if you mirror your vg00 then cut it in half (or just doulble instead of multiplying by 4)

Free PE X PE Size / MIRROR = 856mb free total for vg00.

This is of course unallocated space.

You may have more free space that is already allocated within the filesystem.

For example, you may have /usr @ 60%.

[Scott Van Kalken]

Striping is slightly different, I find it useful to look at both the vgdisplay and an lvdisplay -v /dev/vg?? to figure this out.

If you're using an Autoraid 12h (I think one of your earlier posts alluded to this) you can also use the arraydsp commands. I can't remember the syntax exactly, but man arraydsp should help.

arraydsp -i shows the array ID(s).

I'm a bit rusty, but will try to dig up an arraydsp script I wrote a while ago.

Scott.

[Tom Geudens]

Hi,
Free Space
In vg00 : 1712Mb (4Mb PE size * 428 free PE's)
In vg01 : 8220Mb
In vg02 : 7328Mb
In vg03 : 64Mb

Ok, I'm assuming that you allocated ALL disks to a volume group (you can verify this by doing an "ioscan -fnC disk" and checking that
all these disks are in a "vgdisplay -v").

You cannot just "destroy" a volumegroup. However, it would seem that there's plenty of free space to move things around. You could
(for example) copy all the logical volumes
from vg03 to vg01 and thus free up the disk
used in vg03. You can also use the "pvmove"
utility to free up a disk in a volumegroup
(vg01 and vg02 would be the candidates there
as there's probably more than a whole disk
free in those volumegroups). Note that you can
only use pvmove WITHIN a volumegroup.

So ... I hope this info helps. If not ... or
if you would like specific commands ... just ask :-)

Tom Geudens

[Simon M Appleby]

Guys,

Thanks for the info, a great help all round.

However, I am running from the sublime to the ridlicous in this company. I have been told that the administration of the machine at the Unix level is to be past to a complete (well you can guess). Anyway I do not have access to run would you believe "ioscan" this in the hands of this person AGGHHHH!. Therefore can I ask if there is a script that does the simular to this ?.

In fact reviewing the layers I have noticed that they have balanced the AutoRaid 12H arcoss 2 controllers, with each being allocated a VG, and then each VG having a 2 LUN's (in this Case) of unknown sizes, with then alloctaed LVG's assigned to each LUN. Now I am lead to believe that there is spare space still on the LUN, so allowing (Hopefully) an increase in the LVG's, but again I need to see or understand how to interputt the output from these various aids/tools to see what I have spare either in the LUN or unallocated in the array that I can use to reallocate to the LVG within the LUN.
I hope that makes sence. ;-)

[Tom Geudens]

Hi Simon,
Wow ... this is getting difficult. Absolutely
NO ioscan ? Well, maybe there's hope. Do you have the Support Tools running ? Try the following :
#cstm
#cstm>map
... this should give you the list of ALL system devices, including the arrays. Note down the hardwarepaths of these.
Next, do a "vgdisplay -v" of every volumegroup. Note down the devicefiles (at the end of the list).
Next, do an "lssf" for every one of those devicefiles. This should give you the hardwarepaths.

Ok, you now have a list of the hardwarepaths (the LUN's) and you know which ones are used in the volumegroups. You can get more info on those not used by using "diskinfo".

Hope this helps (I must admire your patience ... I myself would be hitting someone over the head with a very heavy manual by now :-),

Tom Geudens

[Scott Van Kalken]

Simon,

this is where the ioscan -fnC disk comes in handy.

If you notice the device numbers, they will be like cxtydz

that's Card x
Target y
Disk z

This is how you can tell from the combination of arraydsp and ioscan what is connected to what.

I'm guessing that since there are two controllers that the LUNS are alternate pathed.

i.e. if one controller dies, the other can take over.

[Simon M Appleby]

Tom,

Certainly is getting harder, as for the "hitting the bloke over the head with a big book", well he has them now and at a different site so it makes it impossible, however there is a turse e-mail via my boss heading his way.

As for your replies I shall give this a go.

Thanks :)

[Simon M Appleby]

Tom,

When you mention about "Note down the devicefiles (at the end of the list). " When you mean "devicefiles" are you refering to the Physical Volumes the PV Name ?.

:)

[Tom Geudens]

Simon,
I mean devicefiles like
/dev/dsk/c1t6d0

You can then use those as input for lssf
#lssf /dev/dsk/c1t6d0
sdisk card instance 1 SCSI target 6 SCSI LUN 0 section 0 at address 0/0/2/0.6.0
/dev/dsk/c1t6d0

Thus finding the hardwarepath ...
Tom :-)

[Simon M Appleby]

Tom,

Sorry to bother you once again.

I have tried the methods you have advised, and with wonderious results. however, I have a query that hopefully you can help with.

The machine concerned has a RaidArray and external Disk bank. However, reading the output from "lssf" I fine that the ext are included in the LUN's. I have been advised that the LUNs are ONLY for the RaidArray and can not be included, is this true ?. Is there a method to allow this ?.

I'll attach the print outs for you to look over.

[Simon M Appleby]

Tom the 2nd print out, the attachments will only except 1 attachment.

[Tom Geudens]

Simon,
I don't like to admit it :-(, but I don't know ... You've completely exhausted my knowledge of disks and diskutilities and since I don't want to put you on a wrong track ... I suggest you put that question in a new post on the forums. Sorry.

One more hint on the use of cstm. If you would like more information on (for example) your disk array.
cstm>map
cstm>select device 15
cstm>info
cstm>wait
cstm>il

Hope that at least that helps,
Tom

[Simon M Appleby]

Tom,

Thanks for all your help in this, I have infact posted the question. Do you know of any docs on how the cstm works, although the help gives some indication to cmds not functionality is obvious.

[Tom Geudens]

Simon,
http://www.docs.hp.com/hpux/onlinedocs/diag/stm/stm_qik.pdf
A printout of this is one of the few things that a actually keep on paper.

Tom