MSA-1000 - Adding additional SCSI drives

[Ken USA]

Hi - we have a two node (both Alpha running OpenVMS 8.2) cluster connected to a MSA 1000. MSA 1000 had two 36.4GB Ultra-320 SCSIs (one on each side). Today we added eight 72.8 Ultra-320 SCSIs (four on each side)- after bringing both nodes down and turning off MSA 1000. After connecting the new drives, we brought the MSA 1000 up. Few minutes later, we tried to boot one of the nodes.

The system is in a loop (i.e. attempts to boot, crashes, attempts again, crashes ...).

Here's what we see on the console.

BUGCHECK CODE 0000019C
INCONSTATE - Inconsistemnt I/O data base
Crash CPU 00 Primary CPU 00 Active CPUs 00000001
Current Process SWAPPER
Current PSBId 00000001
compressed selective memory dump

boot dga100.1001.0.1.1 -flags 1,0
failed to send read to dga100.1001.0.1.1
failed to read dga100.1001.0.1.1
bootstrap failure

boot dgb100.1002.0.2.1 -flags 1,0
block 0 of dgb100.1002.0.2.1 is a valid boot block
reading 1168 blocks from dgb100.1002.0.2.1

bootstrap code read in
... ... ...
jumping to boot strap code

BUGCHECK CODE 0000019C

Sure we missed a few steps after connecting the drives and before bringing the system up. Any help will be highly appreciated.

[Jefferson Humber · Accepted solution]

Hi,

Make sure the MSA hasn't lost the connection profiles for the OpenVMS hosts. These must be set to OPENVMS, I have seen these go missing before.

I assume the LCD's on the MSA controllers aren't reporting any issues throughout the MSA boot cycle ?

Post a copy of SHOW TECH_SUPPORT from the MSA CLI.

I assume you are running at v4.48 or higher firmware on the MSA.

Hope this helps,

Jeff

[Ken USA]

Both LCDs on the MSA controllers read "SanSwitch 2/8 Ip Adr xx.77.77.77"

[Hoff]

Do you have the current SCSI ECO for OpenVMS Alpha V8.2 loaded?

Try to deconfigure the disks, retry to OpenVMS boot, and -- if you can -- apply the mandatory ECOs and particularly the FC SCSI kit -- and then try reconfiguring the MSA1000 drives.

ftp://ftp.itrc.hp.com/openvms_patches/alpha/V8.2/VMS82A_FIBRE_SCSI-V0300.txt
ftp://ftp.itrc.hp.com/openvms_patches/alpha/V8.2/VMS82A_UPDATE-V0500.txt

There are various crashes fixed here.

[Ken USA]

Here's what we get:

CLI> show version
Firmware Ver 4.48 Build 342
Hardware Dev 7 [AutoRev 0x010000]
Internal EMU Rev 1.86 (SGM06108Q3)

CLI> show disks
Name Box,Bay Bus,Id Size Speed Units
DISK101 1,01 0,00 36.4GB 160 MB/s 0
DISK102 1,02 0,01 72.8GB 160 MB/s none
DISK103 1,03 0,02 72.8GB 160 MB/s none
DISK104 1,04 0,03 72.8GB 160 MB/s none
DISK105 1,05 0,04 72.8GB 160 MB/s none
DISK108 1,08 1,00 36.4GB 160 MB/s 0
DISK109 1,09 1,01 72.8GB 160 MB/s none
DISK110 1,10 1,02 72.8GB 160 MB/s none
DISK111 1,11 1,03 72.8GB 160 MB/s none
DISK112 1,12 1,04 72.8GB 160 MB/s none

CLI> show tech_support
This had a big output and also lots of unreadable characters (bold-reversed question mark - must be some compose characters).

How do I set up the connection profile to OpenVMS?

Thanks.

[Jefferson Humber]

OK, the LCD display is expected. The IP address of the internal switches is the last thing is usually displays. Have a quick cycle through the previous messages with the up/down cursors. See if any errors are highlighted.

The CLI manual can be found here -> http://h20000.www2.hp.com/bizsupport/TechSupport/CoreRedirect.jsp?redirectReason=DocIndexPDF&prodSeriesId=377751&targetPage=http%3A%2F%2Fh20000.www2.hp.com%2Fbc%2Fdocs%2Fsupport%2FSupportManual%2Fc00683579%2Fc00683579.pdf

Check your existing connections with a SHOW CONNECTIONS from the CLI prompt, they should have OPENVMS against them as the profile type.

Please attach the SHOW TECH_SUPPORT output to the thread as file, it would help to rule out any MSA mis-configuration.

Jeff

[Ken USA]

Here's what show connections shows.
CLI> show connections

Name Host-WWNN Host-WWPN Profile U-offset
APGA 20000000- 10000000- Default 0
C9527151 C9527151

APGB 20000000- 10000000- Default 0
C9526C74 C9526C74

BPGA 10000000- 10000000- Default 0
C9526E8D C9526E8D

BPGB 10000000- 10000000- Default 0
C95270DB C95270DB

Changed the prifle to OpenVMS using

CLI> set connection APGA profile=OPENVMS

Entered the above command for APGB, BPGA & BPGB too.

Restarted MSA1000 and then tried to bring one of the nodes up. Still same looping problem.

I am not able to capture output from "show tech_support" as I see lots of bold-reversed question mark signs that messes up the display.

What I was able to write down from clean display is in the text file attached.

Any help or suggestions will be highly appreciated.

Thanks.

[Ken USA]

Not sure what this means, but there is a green blinking light in Enclosure Status Indicator just above the green light for power.

The sign looks somewhat like this (in a circle) - (-/\/\-)

[Ken USA]

CLI> show connections
now shows an extra line giving controller and port info which it did not before.

Connection Name: APGA
Controller 2 Port 1 Status = Online

Connection Name: APGB
Controller 1 Port 1 Status = Online

Connection Name: BPGA
Controller 1 Port 1 Status = Online

Connection Name: BPGB
Controller 2 Port 1 Status = Online

[Bill Hall]

Your cli "display" problems may be due to communication configuration mis-match. I noticed your output from "show other" indicates the terminal is 9600. I seem to recall that the MSA's default is 19200, 8bit, no parity.

Verify your terminal settings and the MSA console settings are consistant.

Bill

[EdgarZamora]

I was wondering why your SHOW CONNECTIONS didn't show the online status. I figured it might've been because of your garbled display. Are you using ACLs? Can you also do SHOW ACL, SHOW UNITS, SHOW THIS_CONTROLLER? Might be helpful to people trying to help you.

[Ken USA]

The output from
CLI> show acl
CLI> show units
CLI> show this_controller
is in the attached file.

Terminal setting was 9,600 and so was MSA setting. Changed MSA TO 19,200. Had to change terminal to 19,200. The result was the same display problem. I am using a VT-420 as a console.

Thank you all for your continued help. I appreciate it a lot.

-Ken

[Jim Lahman_1]

WHat if take out those disks, reboot one of the nodes from the installation CD and apply the patches?

Then, try again.

Attached are my notes (word document) from when I installed and configured disks on my MSA.

I hope this helps you.

Jim

[EdgarZamora]

I'm not sure if this is required, but in your SHOW UNITS listing I don't see a unit identifier. I was always told that I had to set a unit identifier and I always did. I don't know if not having one would cause a problem. You can give it a shot, add a unit identfier with a SET UNIT_ID command. You can get help on it in the CLI.

Good luck

[EdgarZamora]

I just also noticed your controller(s) doesn't show a controller ID. I always thought this was required and had to be unique throughout the SAN. This gives you the $n$GGAnnn number on your VMS system.

[Ken USA]

Just did the following:

CLI> add unit 1 raid_level=0 data=DISK102-DISK105

CLI> add unit 2 raid_level=0 data=DISK109-DISK112

I will now do the following to give these units $1$nnn name.

CLI> set unit_id 1 200 !for $1$DGA200
CLI> set unit_id 2 300 !for $1$DGA300

When we shutdown and restarted MSA1000 yesterday, we lost the device name (it was $1$DGA100) for unit 0. Would entering the following command cause any problem (we have OS + all files created on this as this was the only device we had prior to the upgrade).

CLI> set unit_id 0 100 !for $1$DGA100

Thanks.

-Ken

[Jim Lahman_1]

Don't forget to tell VMS about these disks.

Jim

[Jefferson Humber]

Correct, it's the unit ID of zero that OpenVMS doesn't like. I think that's documented somewhere ? Personally have set my system disk as 10.

You'll need to set the Alpha through WWIDMGR to boot off the LUN from the console prompt.

I have a vague memory that one of my clusters lost the unit ID once as well... Hmmm ?

Jeff

[Robert Brooks_1]

The system is in a loop (i.e. attempts to boot, crashes, attempts again, crashes ...).

Here's what we see on the console.

----
Once you get the system up, post the first page of output from the command CLUE CRASH, within SDA.

My guess is that the previous suggestions you've received are correct, but I'd like to see where it's crashing with INCONSTATE


-- Rob

[EdgarZamora]

And by the way, the FIBRE SCSI patch that Hoff mentioned does contain a fix for an INCONSTATE bug. Not sure if it's for the same one though. But still, it's usually goodness to install the FIBRE SCSI patches.

[Ken USA]

Hi all-

The problem is fixed. I appreciate everyone's help.

I am in a process of writing a detailed (not too big though) report as to what the problem was and what was done to fix it.

Moderator - I will have my report entered as an attachemnt (notepad document) by 7-Mar-2007:20:00 GMT. Once it is posted, you may close this thread.

Special thanks to Jeff Humber, Bill Hall, Edgar Zamara, Jim Lahman, Robert Brooks and Mr. Hoff of Hoffman Labs.

Thank you.

-Ken

[Ken USA]

Hi all-

The attached notepad document describes what the problem was and what was done to fix it.

Thank you all.

-Ken

[Ken USA]

Problem fixed. Closing the thread.