SCSI Errors, cabling problem in HA-Cluster ?

[Christian Schulze]

Hi,

I am setting up a new 2-node Cluster consisting of 2 rp5430 and 2 DS2300

I am doing some tests copying big files (>2G) around and get lots of errors in Syslog:
Aug 21 17:40:05 f699h031 vmunix: lbp->state: 60
Aug 21 17:40:05 f699h031 vmunix: lbp->offset: ffffffff
Aug 21 17:40:05 f699h031 vmunix: lbp->uPhysScript: 95f7e000
Aug 21 17:40:05 f699h031 vmunix: From most recent interrupt:
Aug 21 17:40:05 f699h031 vmunix: ISTAT: 22, SIST0: 06, SIST1: 00, DSTAT: 80, DSPS: 95f7e5c0
Aug 21 17:40:05 f699h031 vmunix: lsp: 0000000000000000
Aug 21 17:40:05 f699h031 vmunix: scb->cdb: 28 00 00 00 00 10 00 00 04 00
Aug 21 17:40:05 f699h031 vmunix: lbolt_at_timeout: 2300048, lbolt_at_start: 2299548
Aug 21 17:40:05 f699h031 vmunix: Script dump [ffffffff95f7e110]:
Aug 21 17:40:05 f699h031 vmunix: 0f000001 95f7e5c0 60000040 00000000
Aug 21 17:40:05 f699h031 vmunix: 9f030000 0000000e 0f000001 95f7e5c1
Aug 21 17:40:05 f699h031 vmunix: LVM: Failed to automatically resync PV 1f051000 error: 5
and so on ...

my Question is, if i did some of the cabling wrong. It is as follows.

each node has 2 A5383 cards with each 2 scsi connectors.
each DS2300 has two BCCs with each 2 scsi-connectors.
Each A5383 connects to one BCC, the 2nd SCSI port on the BCC is terminated.

Do I have to change the SCSI-ID on the A5383s in one Node ?

I am pretty sure that my cabling is right, but ...

Christian

[Jeff Schussele · Accepted solution]

Hi Christian,

Yes - if the A5383 cards are *both* on the same bus, they will need different SCSI IDs. I suspect they're both at 7 & that will be a problem. Change one of them to 6.

Rgds,
Jeff

[Christian Schulze]

OK,
the A5383 are dual SCSI and dual Ethernet cards

ioscan shows the following:
ba 9 0/10/0/0 PCItoPCI CLAIMED BUS_NEXUS PCItoPCI Bridge
ext_bus 4 0/10/0/0/4/0 c720 CLAIMED INTERFACE SCSI C896 Ultra2 Wide LVD
target 7 0/10/0/0/4/0.7 tgt CLAIMED DEVICE
ctl 4 0/10/0/0/4/0.7.0 sctl CLAIMED DEVICE Initiator
/dev/rscsi/c4t7d0
ext_bus 5 0/10/0/0/4/1 c720 CLAIMED INTERFACE SCSI C896 Ultra2 Wide LVD
target 8 0/10/0/0/4/1.0 tgt CLAIMED DEVICE
disk 2 0/10/0/0/4/1.0.0 sdisk CLAIMED DEVICE HP 36.4GMAP3367NC
/dev/dsk/c5t0d0 /dev/rdsk/c5t0d0
target 14 0/10/0/0/4/1.7 tgt CLAIMED DEVICE
ctl 5 0/10/0/0/4/1.7.0 sctl CLAIMED DEVICE Initiator
/dev/rscsi/c5t7d0
target 15 0/10/0/0/4/1.15 tgt CLAIMED DEVICE
unknown -1 0/10/0/0/4/1.15.0 UNCLAIMED UNKNOWN HP A6491A
lan 1 0/10/0/0/6/0 btlan CLAIMED INTERFACE HP A5838A PCI 100Base-TX/SCSI COMBO
/dev/diag/lan1 /dev/ether1 /dev/lan1
lan 2 0/10/0/0/7/0 btlan CLAIMED INTERFACE HP A5838A PCI 100Base-TX/SCSI COMBO
/dev/diag/lan2 /dev/ether2 /dev/lan2

this card number one,
the second is at HW-Path o/12
ba 10 0/12 lba CLAIMED BUS_NEXUS Local PCI Bus Adapter (782)
ba 11 0/12/0/0 PCItoPCI CLAIMED BUS_NEXUS PCItoPCI Bridge
ext_bus 6 0/12/0/0/4/0 c720 CLAIMED INTERFACE SCSI C896 Ultra2 Wide LVD
target 16 0/12/0/0/4/0.7 tgt CLAIMED DEVICE
ctl 6 0/12/0/0/4/0.7.0 sctl CLAIMED DEVICE Initiator
/dev/rscsi/c6t7d0


so, on one node I would change the ID to be like
ctl 6 0/12/0/0/4/0.6.0 sctl CLAIMED DEVICE Initiator
/dev/rscsi/c6t6d0

how can this be done ?

Christian

[Bernhard Mueller]

Christian,

yes the initiator target is always 7 by default, even for the two port cards both have id 7.

So I would suggest to set both to 6 on ONE host.

You need to reboot, halt at BCH, go into ser(vice) menu
enter
scsi id 6
enter scsi to display
you should also set
scsi speed nolimit

on the other node same with id 7

Regards,
Bernhard

[Bernhard Mueller]

Oops,

Christian, your cabling is wrong since most likely you want to mirror one DS2300 onto the other right?

then you need to set full bus mode on the DS2300 and connect BOTH servers to SAME BCC! (the second BCC is only used for failover - or in split bus mode for half of the disks... but I bet you want to see ALL your disks on all nodes, right?)

Regards,
Bernhard

[Christian Schulze]

Well,
I changed the SCSI-ID on both A5383s in Node B,
did not help.
Then I removed all cabling, and set up all again,
et voila, it works.

My cabling looks now like in the Picture.
(Taken from the Manual and edited)

Each second conector on BCC A and B of the DS2300 is terminated.

Christian

[Bernhard Mueller]

Glad you solved it inspite of the wrong information I provided. Of course full bus mode means single bus (with both BCCs) ....

Regards,
Bernhard

[Aaron Barrett_2]

Hi,

I am going to be setting up a similiar configuration and am looking for documentation on clustering two rp5430s.

Thanks,
Aaron