AIX 4.3.3 with two 6228 HBA - Too much rmt created

[fx]

Hello,

On an AIX 4.3.3 server with two 6228 HBAs :

fscsi0 (tapes) - PtP connection to an ADIC snc5100/Scalar1000/LTO1x6
fscsi1 (disk) - connected trough an IBM FC switch to an ESS disk array

Problem is for each physical tape drive, cfgmgr creates two "rmt" devices.
The NetBackup "scan" command shows clearly two "rmt" with the same serial
after cfgmgr has run. "rmdev -d" doesn't help, neither did a clean devices
and a reboot.

A case is going to be open with IBM but knowing it's going to take some
time for misc reasons, I'm curious if this unwanted behaviour may ring
someone's bell here ?

Thanks && amicalement,
fx

Ps: fu2 comp.unix.aix

--
"the only problem with troubleshooting is that
sometimes trouble shoots back" - alt.sysadmin.recovery

[RobH]

"fx" <fx@turlututu.chapo.pointu> wrote in message
news:3f7c22b6$2@ac-versailles.fr...

>
> Hello,
>
> On an AIX 4.3.3 server with two 6228 HBAs :
>
> fscsi0 (tapes) - PtP connection to an ADIC snc5100/Scalar1000/LTO1x6
> fscsi1 (disk) - connected trough an IBM FC switch to an ESS disk

array

>
> Problem is for each physical tape drive, cfgmgr creates two "rmt"

devices.

> The NetBackup "scan" command shows clearly two "rmt" with the same

serial

> after cfgmgr has run. "rmdev -d" doesn't help, neither did a clean

devices

> and a reboot.
>
> A case is going to be open with IBM but knowing it's going to take

some

> time for misc reasons, I'm curious if this unwanted behaviour may ring
> someone's bell here ?
>

You've got two paths to each tape drive over your SAN. The likelyhood is
that each of your tape drives can be "seen" by both FC adapters.

This is probably not what you want if you want to ensure that the tape
drives (which are what?) only use fscsi0 but it can be a resilience
benefit. You need to speak to whoever set up the configuration of the
SAN to change this behaviour if it is not what you want.

Are you getting something similar with your Shark disks?

RobH

[fx]

RobH wrote:

....

> You've got two paths to each tape drive over your SAN. The likelyhood is
> that each of your tape drives can be "seen" by both FC adapters.

I don't think so, as I said, tapes are connected trough a PtP FC link.

-----------
| fcp0 ----(FC 2Gb in PtP mode)------ [snc5100]--(SCSI)--[drives & library]
| AIX box |
| fcp1 ---(FC 2Gb in fabric mode)----|
----------- |
-----------------
| IBM FC switch | -----[ESS disk array]
-----------------

> This is probably not what you want if you want to ensure that the tape
> drives (which are what?) only use fscsi0 but it can be a resilience
> benefit. You need to speak to whoever set up the configuration of the
> SAN to change this behaviour if it is not what you want.

This is a full IBM configuration, so I'd say all drivers come from IBM.
I though maybe the ESS driver for disk multipathing is causing this mess
but according to lsdev outputs, rmt are linked to fcp0 and not fcp1.
This said, I have very little knowledge on ESS configurations.
Another thought is some kind of driver mixup, as both cards are 6228 and
all configuration is software based in the driver (no lputil as with a plain
Emulex card). I remember old HBA driver problem when using same card brand
in the one machine broke some things because of the driver messing up between
the two cards instances.

> Are you getting something similar with your Shark disks?

The array is already in production, so it should be ok - I did not see the
array guys.

Thanks,
fx

[Chris Mattern]

fx wrote:

> RobH wrote:
>
> ...
>

>> You've got two paths to each tape drive over your SAN. The likelyhood is
>> that each of your tape drives can be "seen" by both FC adapters.

>
>
> I don't think so, as I said, tapes are connected trough a PtP FC link.
>
> -----------
> | fcp0 ----(FC 2Gb in PtP mode)------ [snc5100]--(SCSI)--[drives & library]
> | AIX box |
> | fcp1 ---(FC 2Gb in fabric mode)----|
> ----------- |
> -----------------
> | IBM FC switch | -----[ESS disk array]
> -----------------
>

Is the PtP link to the tape drives the *only* connection existing to the
tape drives? If the tape drives are also connected to the switch, you
may be connecting through the switch as well by accident. What are the
location codes for the rmt devices? One should have the location code for
fcp0 in it. If the other has the location code for fcp1 in it, then
you're connecting through the switch as well; fix your zoning so that
the AIX box can't see the tape drives through the switch.

Chris mattern

[fx]

Chris Mattern wrote:

....

> Is the PtP link to the tape drives the *only* connection existing to the
> tape drives?

Yes. That's why I'm scratching my head and suspect some kind of
driver mess.

....

> What are the location codes for the rmt devices? One should have the location
> code for fcp0 in it. If the other has the location code for fcp1 in it, then
> you're connecting through the switch as well; fix your zoning so that
> the AIX box can't see the tape drives through the switch.

According to lsdev output, tapes parent is the fcp0 interface. fcp1 (the fabric
switch one) doesn't shows up in lsdev output - unless lsdev is not enough to
confirm?

Thanks.

--
Three things are certain:
Death, taxes and lost data
Guess which has occurred.

[Chris Mattern]

fx wrote:

> Chris Mattern wrote:
>
> ...
>

>> Is the PtP link to the tape drives the *only* connection existing to the
>> tape drives?

>
>
> Yes. That's why I'm scratching my head and suspect some kind of
> driver mess.
>
> ...

> > What are the location codes for the rmt devices? One should have the

> location

> > code for fcp0 in it. If the other has the location code for fcp1 in

> it, then
>

>> you're connecting through the switch as well; fix your zoning so that
>> the AIX box can't see the tape drives through the switch.

>
>
> According to lsdev output, tapes parent is the fcp0 interface. fcp1 (the fabric
> switch one) doesn't shows up in lsdev output - unless lsdev is not enough to
> confirm?
>

Yeah, at this point it's either a bug or some kind of misconfiguration
in the tape drives themselves, and I don't know anything about the drives
you're using.

Chris Mattern