[Info-vax] removing a node completely

[Stephen Hoffman]

On 2013-02-02 22:17:29 +0000, Phillip Helbig---undress to reply said:

> Just to be clear, neither LANCP> LIST NODE nor SHOW NODE shows the old
> node.  As expected, the new node shows up in SHOW NODE on the one node
> configured as a boot server for it.  What surprises me is that LIST NODE
> shows the new node on all nodes.  Where is the permanent LAN database
> and how did it become cluster-wide?

If you're using DECnet MOP to boot satellites, use CLUSTER_CONFIG.  If 
the LANCP, use CLUSTER_CONFIG_LAN.

If you're using multiple disjoint boot server nodes, you're going to 
have to copy definitions around manually, or reference a common 
database.  I would not start with this configuration.

> The only place I see the old node in LANCP is SHOW DLL, in the two nodes
> which are not configured as boot servers for the old node.  I understand
> what the message is saying:
> 
>    The configuration poller discovered a remote computer with
>    SCSSYSTEMID and/or SCSNODE equal to that of another computer to which
>    a virtual circuit is already open.
> 
> It is clear that this is not a good thing.  So I suppose the question
> is whether it is possible to close this virtual circuit without a
> reboot.

SCSSYSTEMID and SCSNODE are treated as a pair, and remembered as a 
pair.  Once you've booted with a pair and joined a cluster, the pair is 
remembered until the cluster is rebooted.  Get the misconfigured 
cluster member shut down, and ensure the SCSSYSTEMID and SCSNODE pairs 
are set uniquely.  If you want to "break" an existing SCSSYSTEMID and 
SCSNODE pair — as opposed to using a wholly new pair for the host — 
you'll have to reboot the whole cluster, as the cluster remembers all 
of the pairs it has ever seen.


-- 
Pure Personal Opinion | HoffmanLabs LLC