Decserver 700 Crashes VMS Systems?

[Susan Sweeney]

I had a Decserver 700 fail and susequently, all VMS nodes connected to the server crashed. Maybe it broadcasted a break command from the terminal server to each system or overloaded the systems somehow, like a virus. The terminal server displayed an "8" and I had to swap it out. Is there anyway to ensure a bad terminal server does not broadcast garbage back to my systems in such an event? Below is the configuration of the replacement server and one of the ports. I use it for the CA (fomerly polycenter) Console Manager connections to my systems for remote management.

Thanks,

Susan

DECserver 700-16 V1.1A BL45-12 LAT V5.1 ROM V4.0-0 Uptime: 3 02:29:52

Address: 08-00-2B-3C-9E-08 Name: COM1_SERVER Number: 0

Identification:

Circuit Timer: 80 Password Limit: 3
Console Port: 1 Prompt: COM1_SERVER>
Inactivity Timer: 30 Queue Limit: 100
Keepalive Timer: 20 Retransmit Limit: 8
Multicast Timer: 30 Session Limit: 64
Node Limit: 200 Software: PHX700V11

Service Groups: 0

Enabled Characteristics:
Announcements, Broadcast, Dump, Lock

TSM_SVR_COM1_SERVER> sho port 2


Port 2: (Remote) Server: COM1_SERVER

Character Size: 8 Input Speed: 9600
Flow Control: XON Output Speed: 9600
Parity: None Signal Control: Disabled
Stop Bits: Dynamic Signal Select: CTS-DSR-RTS-DTR

Access: Remote Local Switch: None
Backwards Switch: None Name: PAPAGO_P2
Break: Remote Session Limit: 1
Forwards Switch: None Type: Ansi
Default Protocol: LAT

Preferred Service: None

Authorized Groups: 0
(Current) Groups: 0

Enabled Characteristics:
Failover, Input Flow Control, Lock, Loss Notification, Message Codes,
Output Flow Control, Verification

[Vladimir Fabecic]

Hello
It is very interesting that terminal server crashes VMS host. I would like to see ANALYZE/CRASH_DUMP to see what really crashed systems. How to check that bad terminal server does not broadcast garbage back to your systems? You can use TCPTRACE utility. But I would do the following: I would use a hub (repeater, not switch)) and connect terminal server and Linux host (laptop) to it. Hub should be connected to existing network switch. Then you watch the traffic using tcpdump from Linux. This is the wayhow I fixed problem with printer server.

[Volker Halle]

Susan,

let me try to first clarify the VMS crash aspect of your question: did the VMS nodes really crash (and write a crash dump) or did they just stop responding and were left at the console prompt (>>>).

If crashdumps have been written, you should find crash errlog entries and - even better - also CLUE files with the crash footprint (on Alpha: CLUE$COLLECT:CLUE$node_ddmmyy_hhmm.LIS, on VAX: CLUE$OUTPUT:CLUE$LAST_node.LIS). We could then try to work out from the crash information, what has happenend.

If the VMS nodes were forced to the console prompt (>>>), most probably by receiving a signal, there may be ways to disable BREAK processing on the console depending on your type of system. See the VMS FAQ chapter 14.17

http://h71000.www7.hp.com/faq/vmsfaq.txt

Volker.

PS: I'm trying to post this reply since about 6 hours, but ITRC kept failing with 'Error while posting the reply'

[Susan Sweeney]

Volker and Vladamir are right. The systems were all halted like they received a BREAK. They did not crash. All sysems were sitting at >>>.

The 14.17 FAQ matches my situation. If I do disable the BREAK, how can I remotely halt my systems if need be? Would the only way be to press a button?

Of course, choosing a "surprise" halt over the times when I need to "force" a halt, I would rather disable the surprise halt and worry about how to get to the systems physically if there is a need to halt them.

Thanks for your help!!

Susan

[Volker Halle]

Susan,

with Console Manager, I usually halt the systems by pressing

To allow a to get through to the console port of your system, you might have to change the default BREAK character in the Console Manager, otherwise, you'll get the following prompt, if you type after $ CONSOLE CONNECT node :

Do you want to transmit a break? [Y/N]

If you answer yes, your will be converted to a and if you had disabled on the console, it won't work to halt the system.

To allow a to get throught, I use $ CONSOLE CONNECT/BREAK=CTRL_D node - this would transparently let get through. If you accidently type CTRL-D (less typical keystroke for VMS system managers), you'll get the above prompt and reply N and nothing happens.

As said, the abilitya and commands to disable BREAK on the console depends on your type of systems.

Volker.

[Lawrence Czlapinski]

Susan, the systems received breaks (when the terminal server went down) on the console lines which caused them to halt the systems.
Unfortunately, if you disable break on the systems you wouldn't be able to remotely break from the console manager. However, you could remotely access, fix and/or crash the system if needed using Availibilty Manager.
Lawrence

[John Lapchuk]

The outage you experienced was caused by the software version you are running on the DecServer. Version 1.1 has known (to HP) issues and HP suggests version 1.3 or better. Upgrade your DS700 software and problems will go away...

JohnL

[Susan Sweeney]

John,

I scanned my latest software product library and no sign of Decserver software. I can go search the website for it, but if you know where I can get it that would help expedite the process.

Thanks,

Susan

[Volker Halle]

Susan,

the latest release of the DECserver 700 software (UPI: XA5 SPD 38.73) is available on the Layered Products CDs is V1.1C (saveset name DS7C011 on the Q2/2003 Alpha LP CDs).

The DECserver business has been transferred to Digital Networks many years ago (www.dnpg.com).

There is also DECserver Network Access Software (DNAS) (UPI: 0LW), which also runs on DECserver 700, but is not and seems to have never been on the OpenVMS LP CDs.

I've also found a NAECO02022 ECO kit for the DNS software on the HP FTP site, but it's access restricted.

Volker.

[Susan Sweeney]

Volker,

Thanks. I looked and looked for 1.3 as John had suggested but with no luck. I wonder where he got the 1.3 information.

I will have to go to storage to dig out Q2/2003. Will DS7C011 provide anything more than DS7A011? By the way, I have not tried to change the BREAK command yet :), nor have I had time to look at Availability Manager as Lawerence suggested. I was thinking I could upgrade the software if it will help.

Regards,

Susan