Connecting a Disk System 2120 to two rx2660s

[GBR]

Hi,

I'm connecting two 2120 Disk Systems (Field Racks) to two rx2660s (11i v3), so the disks will eventually be shared between the servers, in case of a failover via ServiceGuard. Each Disk System is holding 4 73 Gig drives each. Both the servers and the disk systems are brand new, so I'm assuming at this point that there is no malfunctioning hardware.

When I connect the cables and boot up the servers I see the following errors in the syslogs on both servers:

SCSI Ultra320 0/2/1/0 instance 0: External SCSI bus reset detected. Condition cleared, no intervention required.
SCSI Ultra320 0/2/1/0 instance 0: IO Type : SCSI IO has timed-out. Target ID: 12, LUN ID: 0. Test Unit Ready Command - CDB: 00 00 00 00 00 00
class : lunpath, instance 1
lun path (class = lunpath, instance = 1) belonging to LUN (default minor = 0x1) has gone offline. The lunpath hwpath is 0/2/1/0.0xa.0x0

SCSI Ultra320 0/2/1/0 instance 0: An IO timeout condition was detected. Condition cleared, no intervention required.
SCSI Ultra320 0/2/1/1 instance 1: Driver initiating SCSI bus reset. Condition cleared, no intervention required.
SCSI Ultra320 0/2/1/0 instance 0: Controller reset has been successfully completed.
SCSI Ultra320 0/2/1/0 instance 0: Driver initiating SCSI bus reset. Condition cleared, no intervention required.
SCSI Ultra320 0/2/1/1 instance 1: External SCSI bus reset detected. Condition cleared, no intervention required.

In addition to this at different times when I run an "ioscan -fnC disk" the disks from the Disk Systems are "CLAIMED" and then other times they indicate "NO_HW".... does anyone have suggestions on what my issue is?

Thanks,
GBR

[cnb]

Hi,

Can you be more specific on the SCSI cabling and hardware connections. It sounds like h/w contention but more info is needed on the complete configuration topology.

What happens if you only have one server up. Do you still see errors?

thx,

[Torsten. · Accepted solution]

When sharing SCSI disks between 2 servers you MUST set the ID of one HBA from 7 to 6!

Also make sure none of your disks is using ID 6!

Otherwise you have a SCSI ID conflict.

[Torsten.]

Example:

# mptconfig -i 6 /dev/mpt10

Confirm: Do you want to change scsi parameters (y/[n]): y

Scan For Devices to avoid id conflict ... [Passed]
Status .................................. [Successful]

[GBR]

So server 1 connects to disk system A using SCSI ID = 7, and server 2 connects to disk system A using SCSI ID = 6? Is this what you mean?

[Steven Schweda]

> So server 1 connects to disk system A using
> SCSI ID = 7, [...]

I wouldn't say "connects to [...] using".

A SCSI adapter has a SCSI ID, just like every
other device on a SCSI bus. Most commonly,
its SCSI ID is 7, but every device on the bus
needs its own SCSI ID, so if you connect two
SCSI adapters to the same bus, then they
can't both have ID 7. 6 is different from 7,
so it would be ok, assuming that no other
device (disk, say) is set to ID 6.


> [...] does anyone have suggestions on what
> my issue is?

If _you_ can't keep track of your offspring,
how should _we_ know? Or did you mean
_problem_?

[GBR]

Two things:

1. In order to resolve this, the SCSI ID on the second host was changed from the default 7 to 6, and this got us to the next step.... but this led to different problems, more conflicts.

2. To solve the next problem we switched the 2120 ID switch at the back from position A (0,2,4,6), to C (9,11,13,15), which resolve the fact that one disk has an ID of 6 conflicting with toggle position A.

Everything is working.

Thanks Torsten.