Shared SCSI tape - Ultrium2 works, Ultrium3 doesn't?

[sgadsby_1]

I have a couple of sites where we have set up a shared SCSI channel via Adaptec SCSI adapters with different IDs. Either end of the bus is terminated at each of two servers that form an MSCS 2-node cluster.

This works great for Ultrium 2 tape drives,
however it does not work for Ultrium 3 tapes!

Any idea why?

If I connect both an Ultrium2 and an Ultium3 onto the same bus, then:
* I can successfully read and write to the Ultrium2 from either server.
* I cannot read at all from the Ultrium3, from either server.

I think the termination is ok (SCSI cards have jumper for hardware termination - works even if server is offline).

So I wondering if there is a difference in the tape drives from generation 2 to 3 that you know of that might be causing my grief?

Thx for any insight,
Simon.

[Marino Meloni_1]

HP recommend to attach only One LTO3 per bus, this because one LTO could saturate the bandwith, and having more drive is useless.
Also when you have a LTO3 on the bus, you should have the terminator on that drive, put as last component of the scsi chain

in any case, having a LTO3 working on that bus, the other device will usualy generate time out if you try to use both, this is the reason why only one LTO3 per scsi bus is allowed

[sgadsby_1]

Thanks Marino.

So I wonder if the higher bandwidth consumption of an LTO3 means that it has different requirements for termination, or something like that?

FYI we only have one LTO3 connected.
* If there is LTO2 + LTO3, then the LTO2 works fine and the LTO3 fails.
* If there is only an LTO2, then it works fine.
* If there is only an LTO3, then it doesn't work at all.

The HP Storageworks software complains about errors "consistent with" cabling or termination problems, however I suspect it must be something else.

[Marino Meloni_1]

LTO3 specification are more strict, so you should have correct cable and terminator, this is a must.

then LTO3 should be the only tape device on the scsi chain

and the LTO3 should be the last of the scsi chain, with the terminator on it (in case you are connecting a library controller on the same scsi bus)

Now in your case, "if there is only a lto3, then it doesn't work at all" this should work, if not, I suspect come cabling problems, like quality or specification of cable and terminator, or also a bent/broken pin on one of these two components

[sgadsby_1]

> Now in your case, "if there is only a lto3, then it doesn't work at all"
> this should work, if not, I suspect come cabling problems, like quality or
> specification of cable and terminator, or also a bent/broken pin on one of
> these two components

The tape drive works fine if it is on a bus that is connected to a single server, with a terminator on the end of the tape.

However when I have it set up as:
SVRA <--> LTO3 <--> SVRB
it does not work, with the same symptoms from either end.

What I see actually is apparently successful write, but failure to read any data (therefore the verify fails).

[Marino Meloni_1]

ok now i understand your configuration.

HP do not support this multiiniziator configuration via scsi on stand alone tapes, so I cannot tell you why it work with LTO2 and not with LTO3, may be the firmware get confused by the two iniziators, or the problem occur due to the cabling.

Be sure to have activated the termination on both HBA (not always they are active)

I can suggest you to slow down the the HBA speed and see if the problem still occur.

I have never seen this config anyway, usualy shared drives are over the FC via a Network Storage Router that support multiInitiator environment

[sgadsby_1]

Yes, everyone seems to say that they don't support it, but none-the-less it should work! And it does work very well with LTO2. We are using Backup Exec, and the documentation provides several scenarios for shared SCSI over copper tape storage.

Both SCSI Adaptors are terminated physically with a jumper on the card. The adapter doco says this is required if you have another server on the other end of the scsi bus as we do - so there's another party saying that it should work ok!

I doubt there is any conflict for having two scsi initiators, because the scsi channel is only accessed from one node at a time. If one is active (ie. Backup Exec is running on it) then the other node does not touch it at all.

Good suggestion re speed - We did try limiting it to 40MB/s and also switching off a few SCSI options on the adapter, but with the same results.

I can only wonder that you may be onto something with yr comment re tape firmware. Perhaps they have something in there that does a scsi bus scan and gets confused by the presence of two HBAs. Or maybe it is something to do with reflections and termination after all - perhaps the onboard termination is passive?

Sigh!