SEO

Hi,

I have basically two questions (for HP-UX 11.11 and newer):

HP technican said HP-UX can deal with non-contiguous LUNs for disks, but it
seems it cannot; is there a patch maybe? The sympom is: A FC-based disk array
(HP EVA 4000) presents these LUNs to a host: 0,1,2,3,4, 9. When doind an
ioscan on the FC-HBA, device files are created for LUNs 0-4, but not for 9.
That's my question: Is HP-UX able to detect non-contiguous LUNs automatically?

The second question is: If HP-UX cannot detect those automatically, can it
access the hardware if I create a correct device file manually?

Finally, it seems HP-UX can only ioscan a complete FC-HBA, but it cannot
ioscan for a specific LUN. Maybe that means that the answer to both questions
above is "no".

Any insights on that? (The background for this is: We want to present the same
virtual disks as the same LUNs in all nodes of a cluster where the nodes are
not perfectly symmetric).

Regards,
Ulrich

On Jul 26, 2:51 am, Ulrich Windl <Ulrich.Wi...@RZ.Uni-Regensburg.DE>
wrote:
> Hi,
>
> I have basically two questions (for HP-UX 11.11 and newer):
>
> HP technican said HP-UX can deal with non-contiguous LUNs for disks, but it
> seems it cannot; is there a patch maybe? The sympom is: A FC-based disk array
> (HP EVA 4000) presents these LUNs to a host: 0,1,2,3,4, 9. When doind an
> ioscan on the FC-HBA, device files are created for LUNs 0-4, but not for 9.
> That's my question: Is HP-UX able to detect non-contiguous LUNs automatically?

Uhm, have you tried it with LUNs smaller than 7, e.g., 0,1,5,6? If
that works then the issue is probably related to volume set addressing
configuration in the array.

See http://docs.hp.com/en/AB378-96003/ch01s10.html

--
Luis P Caamano
Atlanta, GA USA

lcaamano <lcaamano@gmail.com> writes:

> On Jul 26, 2:51 am, Ulrich Windl <Ulrich.Wi...@RZ.Uni-Regensburg.DE>
> wrote:
> > Hi,
> >
> > I have basically two questions (for HP-UX 11.11 and newer):
> >
> > HP technican said HP-UX can deal with non-contiguous LUNs for disks, but it
> > seems it cannot; is there a patch maybe? The sympom is: A FC-based disk array
> > (HP EVA 4000) presents these LUNs to a host: 0,1,2,3,4, 9. When doind an
> > ioscan on the FC-HBA, device files are created for LUNs 0-4, but not for 9.
> > That's my question: Is HP-UX able to detect non-contiguous LUNs automatically?
>
> Uhm, have you tried it with LUNs smaller than 7, e.g., 0,1,5,6? If
> that works then the issue is probably related to volume set addressing
> configuration in the array.
>
> See http://docs.hp.com/en/AB378-96003/ch01s10.html

Linux (used as a test) finds them all...if that says anything.

>
> --
> Luis P Caamano
> Atlanta, GA USA

"Ulrich Windl" <Ulrich.Windl@RZ.Uni-Regensburg.DE> wrote in message
news:871wevslne.fsf@pc9454.klinik.uni-regensburg.de...
> Hi,
>
> I have basically two questions (for HP-UX 11.11 and newer):
>
> HP technican said HP-UX can deal with non-contiguous LUNs for disks, but
> it
> seems it cannot; is there a patch maybe? The sympom is: A FC-based disk
> array
> (HP EVA 4000) presents these LUNs to a host: 0,1,2,3,4, 9. When doind an
> ioscan on the FC-HBA, device files are created for LUNs 0-4, but not for
> 9.
> That's my question: Is HP-UX able to detect non-contiguous LUNs
> automatically?
>
> The second question is: If HP-UX cannot detect those automatically, can it
> access the hardware if I create a correct device file manually?
>
> Finally, it seems HP-UX can only ioscan a complete FC-HBA, but it cannot
> ioscan for a specific LUN. Maybe that means that the answer to both
> questions
> above is "no".
>
> Any insights on that? (The background for this is: We want to present the
> same
> virtual disks as the same LUNs in all nodes of a cluster where the nodes
> are
> not perfectly symmetric).
>
> Regards,
> Ulrich

I have not (yet) tried this on an EVA, but HP-UX (11.00, 11.11 and 11.23)
will work quite happily on non-contiguous LUNs off a VA array.

Will test on an EVA6000 (with HPUX 11.23 IA) for you.



--
Posted via a free Usenet account from http://www.teranews.com

"Kilgaard" <Kilgaard@hotmail.com> writes:

[...]
> I have not (yet) tried this on an EVA, but HP-UX (11.00, 11.11 and 11.23)
> will work quite happily on non-contiguous LUNs off a VA array.
>
> Will test on an EVA6000 (with HPUX 11.23 IA) for you.

Thanks. I only noticed that for HP-UX the LUNs are not presented as type
"disk", but as "RAID"; maybe that makes the difference? (Linux would detect
the LUNs, but would not use those as disks (until the presentation OS was set
to Linux...))

Ulrich

"Ulrich Windl" <Ulrich.Windl@RZ.Uni-Regensburg.DE> wrote in message
news:87r6mmfhci.fsf@pc9454.klinik.uni-regensburg.de...
> "Kilgaard" <Kilgaard@hotmail.com> writes:
>
> [...]
>> I have not (yet) tried this on an EVA, but HP-UX (11.00, 11.11 and 11.23)
>> will work quite happily on non-contiguous LUNs off a VA array.
>>
>> Will test on an EVA6000 (with HPUX 11.23 IA) for you.
>
> Thanks. I only noticed that for HP-UX the LUNs are not presented as type
> "disk", but as "RAID"; maybe that makes the difference? (Linux would
> detect
> the LUNs, but would not use those as disks (until the presentation OS was
> set
> to Linux...))
>

On my 11.23 IPF server I presented a new LUN6 to the server. This server
currently has LUNs 1 and 3 presented, but it did have a LUN2 at some point
in the past (and hence the gap in the numbering). Note the LUN6 is still on
the same sctl as the existing LUNs. I did not want to polute the server with
addition sctl's so kept the LUN as 7 or less. This may be the issue you are
seeing in that LUN9 will be on a different sctl, but I have another server
(11.00 attached to a VA array) that uses LUNs 100 thru 104.

ioscan (as per the man page) only works at the "bus" level, which
effectively means at the EVA controller level, and not the individual LUN
level. This should not cause any issues, as the scan will only be talking to
the EVA, and should not disrupt traffic to any other devices (like FC
tapes).

# ioscan -fn -H 0/3/1/0.1.22.0.0

Class I H/W Path Driver S/W State H/W Type Description
================================================== =========================
ext_bus 5 0/3/1/0.1.22.0.0 fcd_vbus CLAIMED INTERFACE FCP
Array Interface
target 4 0/3/1/0.1.22.0.0.0 tgt CLAIMED DEVICE
ctl 3 0/3/1/0.1.22.0.0.0.0 sctl CLAIMED DEVICE HP
HSV200
/dev/rscsi/c5t0d0
disk 6 0/3/1/0.1.22.0.0.0.1 sdisk CLAIMED DEVICE HP
HSV200
/dev/dsk/c5t0d1 /dev/rdsk/c5t0d1
disk 11 0/3/1/0.1.22.0.0.0.3 sdisk CLAIMED DEVICE HP
HSV200
/dev/dsk/c5t0d3 /dev/rdsk/c5t0d3
disk 37 0/3/1/0.1.22.0.0.0.6 sdisk CLAIMED DEVICE HP
HSV200


# insf

insf: Installing special files for sdisk instance 37 address
0/3/1/0.1.22.0.0.0.6

All as expected. Note that this server is dual attached (so four paths to
the EVA) but I limited the ioscan to just one path, and it created just the
one special file. The other three paths remain undetected.



--
Posted via a free Usenet account from http://www.teranews.com

"Kilgaard" <Kilgaard@hotmail.com> writes:

> "Ulrich Windl" <Ulrich.Windl@RZ.Uni-Regensburg.DE> wrote in message
> news:87r6mmfhci.fsf@pc9454.klinik.uni-regensburg.de...
> > "Kilgaard" <Kilgaard@hotmail.com> writes:
> >
> > [...]
> >> I have not (yet) tried this on an EVA, but HP-UX (11.00, 11.11 and 11.23)
> >> will work quite happily on non-contiguous LUNs off a VA array.
> >>
> >> Will test on an EVA6000 (with HPUX 11.23 IA) for you.
> >
> > Thanks. I only noticed that for HP-UX the LUNs are not presented as type
> > "disk", but as "RAID"; maybe that makes the difference? (Linux would
> > detect
> > the LUNs, but would not use those as disks (until the presentation OS was
> > set
> > to Linux...))
> >
>
> On my 11.23 IPF server I presented a new LUN6 to the server. This server
> currently has LUNs 1 and 3 presented, but it did have a LUN2 at some point
> in the past (and hence the gap in the numbering). Note the LUN6 is still on
> the same sctl as the existing LUNs. I did not want to polute the server with
> addition sctl's so kept the LUN as 7 or less. This may be the issue you are
> seeing in that LUN9 will be on a different sctl, but I have another server
> (11.00 attached to a VA array) that uses LUNs 100 thru 104.

[...]

Exactly that was the problem: I was looking for some hwpath like *.9 (which
does not exist, because SCSI LUNs are restricted to the range of 0-7). From
the docs I learned that a EVA LUN is presented in a fabric to something like
FC-Adapter, ..., Bus=(LUN / 128), Target=(LUN % 128) / 8, Lun=LUN % 8

Sorry, but the "Lun" was confusing me. Our older storage systems all had LUN
numbers < 7.

Regards,
Ulrich