Conference kernel::csguk_systems

131.0. "LOADSA ********* IN SHOW CLUSTER" by KERNEL::CLARK () Thu Feb 07 1991 14:22

	As you may see from the "Subject" field, I'm trying to find out what 
is the significance of the "******" in the sys_id field of the show-cluster 
display.

	The background to this is, that a field engineer, added an HSC70 to a
configuration which already had 2 x HSC50's.

	When he ran a "show cluster", he got the *** for the two HSC50's and a 
genuine number for the '70.

	He couldn't remember if this had been the situation before adding on 
the '70, so we don't have a feel for whether the add-on actually brought about 
this anomaly.

	In addition, the add-on was done "on-the-fly".

	There were no duplicate numbers set up anywhere.

	Everything else appears to be running OK.

			************************

From:	GOONS::PLANK        "My Mum tried to call me 'Wooden'" 27-JUL-1990 

Subj:	RE: show cluster/add sys_id gives ******

As to your query - at the moment, this is my theory and I have yet to check it 
out.

The SCSSYSTEMID of a node in a cluster is defined in the SCS software layer. 
Every node must know the id of every other node. If you look at the node id of 
all the VAXes you will get the familiar "DECnet area no * 1024 + node within 
area". However, HSCs are not DECnet nodes and so the first HSC in a cluster 
gets a SYSID based on the MAXIMUM area number allowed + the MAXIMUM number 
within an area. ie: AREA = 65
		    NODE = 512
 do the sums on this and you get 65025. As all SYSIDs in a cluster MUST BE 
UNIQUE, the extra HSCs get numbers which go outside this range and can 
therefore not be reported by the cluster software.

I'm trying various contacts but I'm sure all they'll do is put more granularity 
into this description.

I'll mail you more when I know.


Steve.

			********************************

From:	GOONS::PLANK        "My Mum tried to call me 'Wooden'" 27-JUL-1990 

Subj:	RE: show cluster/add sys_id gives ******

Dave,

It's much easier than I said. I just went and had a look at all our HSCs. The
SYSID set on the HSC was E-f***ing-normous. On the HSCs which had high numbers
I got ******** in the display, in the low numbers I got the SYS_ID. Therefore
if you increase the column width of the SYS_ID field in SHO CLU/CON it all
comes out OK. 

Do:	Command> set sys_id/width=xx (I tried 30 and it was OK).

Steve