OVMS Cluster Config

[#1]

Hi All,

I am trying to make a OVMS SCSI Cable Cluster system. Two cluster nodes are Alpha server 4100 and DS-20.
I have attached the the snap for cluster configuration.
I have connected a storage box common to both system, but both the system are not able to read the disk in storage box at the same time.
But if connected one by one both of them can read it.
Can anyone help me what I am doing wrong with the cluster setup.
I am using a 68 pin SCSI cable and every connector is SE.

Regards,
ajaydec

[Volker Halle]

Ajaydec,

did you appropriately set the SCSI adapter ID to different values on both Alpha systems ?

Volker.

[#3]

I don't know how to check and set the SCSI adapter id. Could you help me in this.

[Volker Halle]

ajaydec,

you can check and set the SCSI adapter ID at console level:

>>> show pk*

>>> SET PKx0_HOST_ID n

See Table A6 in the 'Guidelines for OpenVMS Cluster Configurations' manual.

Volker.

[#5]

Volker,

I have set the id of
PKA0_host_id 5
PKB0_host_id 6
PKC0_host_id 8
on Alpha 4100 and

PKA0_host_id 4
PKB0_host_id 7
PKC0_host_id 7
on DS-20 server.

But then also its not working.
Note: I am not able to change host_id of PKB0 and PKC0 on DS-20.

Regards,
ajadec

[Bill Hall]

ajaydec,

Your diagram does not indicate which scsi bus is the shared bus. It also doesn't tell us what model of scsi adapter you are trying to use. The adapters on the shared bus should be either KZPSA or KZPBA. They should also have the terminating resistors removed from those adapters.

Bill

[John Abbott_2]

1. What HBAs from your list above are actually connecting to the cabling for each system ? .e.g 4100=PKA0 DS20=PKB0

2. What is the "storage box" ?

3. is there an actual error here ? from console from vms ? can you post it ?

Personally, I can't see this cabling working ? (maybe I'm reading it wrong).

Keep it simple to start with... For example..

4100 PKx HBA connected to a Y cable - one branch connected to terminator (if the HBA doesn't have one ? (KZPBA-CB?)) and the other leading to the DS20, again Y cable into the DS20 HBA with the other end conencting to your storage. One system with a scsi id of 6 and the other 7.

Regards
John.

[#8]

Hi,

4100 PKA SE SCSI cable is connected to a Y cable - one branch connected to SE terminator and the other leading to storage box ( storage box has L shape having two 68 pin connector).
Again, DS-20 PKB SE SCSI cable is connected to a Y cable - one branch connected to SE terminator and the other leading to storage box.

Regards
ajaydec

[John Abbott_2]

OK, thanks, but....

>> 2. What is the "storage box" ?
>> 3. is there an actual error here ? from console from vms ? can you post it ?

Cheers
John.

[#10]

Hi,

2) Storage box is SCSI storage box. BA350
3) Actually there are no error on the console. One system is booted properly and when I tried to boot the other system then device goes offline and it again mount it, again it goes offline and again it mounts it... it repeats....

Sometime I also see lot of device at >>> prompt after giving show dev command but I have inserted only one device in storage box.

Regards,
-Nitin

[Bob Blunt]

First things first... The 68-pin high density connector is not necessarily an indicator of a single-ended SCSI Adapter. I wish I could say without any reservation that all of the ones I've seen "stamped" with the SE "emblem" have always and forever been correct. Knowing what those SCSI adapters really are, what model number, is key. Some of them, even in the various "families" (SE vs LVD vs HVD) don't play well together. It would be easier if you knew you had, for instance, a KZPBA-CA in both systems because the KZPBA-CA and KZPBA-CB look similar. The KZPSA-BB is a much larger card, it has a 68-pin HD connector and it's never going to work with either KZPBA-Cx card, ever. Not unless you can convert to the right SCSI signalling.

The BA350 doesn't really support the use of a personality module, you probably should be connecting your cables directly to the connectors on the backplane of the shelf. It only supports single ended SCSI. A DWZZA-VA can be installed in slot 0 to adapt to different SCSI interfaces.

In other words, we need more information to help. What type SCSI Adapters, what type (and color) storage shelf, what type personality module (if any), etc. There are many, many parts that looked similar, so specifics are mandatory.

[#12]

Hi All,

In the triplet prompt I can see the disk from both the server as:
From DS-20
DKA600

and From Alpha 4100
DKB600

But when I tried to boot DS-20 from DKA600, I got the following error:

bugcheck code = 0000036C: PROCGONE, Process not in system
Crash CPU: 00000000 Primary CPU: 00000000 Node Name: OVMS83
Supported CPU count: 00000002
Active CPU: 00000000.00000003
Current Process: sysinit
Current PSB ID: 00000001
Image Name: sysinit.exe


Starting compressed selective memory dump at 1-oct-2007 17:19................

...Complete *************

System shutdown complete

Can anyone help me, why this error is coming and what's the solution for this.

Regards,
-ajaydec

[Heinz W Genhart]

Hi ajadec

I had this error on a cluster with two systemdisks, if the labels of both systemdisks was same. Changing the label of one of the systemdisks solved the problem for me.

Regards

Geni

[Volker Halle · Accepted solution]

ajaydec,

if you are booting from the same system disk connected to different adapters (PKA/PKB), you need to use port allocation classes to make the device name unique ! Otherwise OpenVMS will crash when trying to mount the disk from the 2nd system.

Volker.

[#15]

Hi Geni,

I am trying to boot both the system from the same system disk, but I was not able to boot even the single system using that disk.

Regards,
ajaydec

[#16]

Hi Volker,

Can you help me on "how to use port allocation classes to make the device name unique !"

I think that's the problem because of which I am not able to boot the server.

Regards,
ajaydec

[Volker Halle]

ajaydec,

you need to provide consistent and detailled information about the configuration, otherwise we can't help you.

In the beginning, you said: 4100 PKA and DS20 PKB constitute the shared SCSI bus. In another reply you say: boot DS20 DKA600 and boot 4100 from DKB600

When the DS20 crashed during boot, was the 4100 running from the same system disk at that time ? To find the reason for the crash, analyze the dump with $ ANAL/CRASH SYS$SYSTEM:SYSDUMP.DMP and provide the output of SDA> CLUE REGISTER (at least the value of R0).

If you can't configure the shared SCSI bus to use the SAME SCSI controller letter (PKx) at OpenVMS level on both systems, you need to use Port Allocation classes. Use @CLUSTER_CONFIG and the CHANGE menu option.

Volker.

[#18]

Sorry Volker,

I did some mistake before.
Let me clear it again....
Alpha 4100 to PKB (DKB600) and DS-20 to PKA (DKA600).
When I tried to boot DS-20 using command:
>>> boot dka600 -fl 0,0
I got the above problem.

At that time Alpha 4100 is not running.

Regards,
ajaydec

[Volker Halle]

ajaydec,

so the first step is to get the DS20 successfully boot from DKA600.

Can you boot the DS20 from an OpenVMS Operating System CD and look at the system dump on DKA600:

$$$ MOUNT/OVER=ID DKA600:
$$$ ANAL/CRASH DKA600:[SYS0.SYSEXE]SYSDUMP.DMP
SDA> EXA R0
SDA> SHOW DEV D ! do you see DKA600 in the dump ?
SDA> EXIT

Volker.

[#20]

Volker,

I have attached the output.

Regards,
ajaydec

[#21]

Vokler,

I forgot to mention that when I remove the connection of Alpha 4100 from the storage box then I can boot DS-20 from the disk in storage box.

I removed Alpha 4100, booted DS-20 and then executed the command that you asked for.

Regards,
ajaydec

[Volker Halle]

ajaydec,

R0 = 104D8CE4 = %IMGACT-F-SYMVECMIS, shareable image symbol vector table mismatch

This is a software problem footprint: there is some image (most likely in the system-specific root of this node [SYS0.SYSLIB]), which causes this problem. You should see the SAME crash, if you try to boot the 4100 from that root. If you cannot reproduce the crash on the 4100, it could also be caused by some hardware-problem in the disk access path, which would corrupt data read from the disk...

Volker.

[Volker Halle]

ajaydec,


I have set the id of
PKA0_host_id 5
PKB0_host_id 6
PKC0_host_id 8
on Alpha 4100


If this is still TRUE, you have set the SCSI ID of the shared SCSI bus host adapter on the 4100 to the SAME VALUE as the disk you're trying to boot from. Not a brilliant idea ;-(

Make sure the SCSI IDs are UNIQUE on the shared SCSI bus. You would typically set the HOST adapters to 6 and 7 and use the lower SCSI-IDs for the disks.

Volker.

[#24]

Thanks to all for your time and special thanks to Volker.

The problem got solved by just removing the two SE terminator which I attached to both the systems. I don't know why but they were causing some hardware problem as mentioned by Volker.

After that I match the alloclass for DS-20 with that of Alpha 4100.

i.e. on DS-20 PKA is 5
Alpha 4100 PKB is 5.

Regards,
ajaydec

[#25]

regards,
ajaydec