Map to cluster volume on Linux (ncp)

[James Taylor]

If I am understanding what you are saying, you are trying to migrate resources independent of the cluster virtual servers that they are associated with.  This would be invalid. resources have unique virtual cluster server objects that they are associated with, and the cluster virtual server objects are migrated to/from or onlined/offlined on a cluster node (host) as a unit. When you migrate R1 to N2, then VNCP1 will be on N1. R1 should never have any direct connection with VNCP2.
If R1 and R2 happen to be on the same node, then you will be able to see all of the volumes associated with each resource from each virtual cluster server and the the node they are on, but you should only access the resources via the virtual cluster server IP address.  If you do it any other way, then connections will fail when you migrate the resources.  Cluster resources should never be accessed (except for specific administration purposes) by any other method than the virtual cluster server address.

-jt
 

James Taylor
The East Cobb Group, Inc.
678-697-9420
james.taylor at eastcobbgroup.com
http://www.eastcobbgroup.com




>>> "joea at j4computers.com" <joea at j4computers.com> 5/15/2009 03:21 PM >>> 
One last, (I swear) observation. 

I create two VNCP objects, VNCP1 and VNCP2. 

Load Resources R1 and R1 on node N1. Have bind statements like this: 

(R1 ) - "ncpcon bind --ncpservername=VNCP1 --ipaddress=R1_IP_address" 
(R2 ) - "ncpcon bind --ncpservername=VNCP2 --ipaddress=R2_IP_address" 

The volumes associated with each resource show up under VNCP1, VNCP2 is empty. 
Migrate R1 to node N2. The volume associated with R1 moves to VNCP2. 
Migrate R2 to node N2. The volume associated with R2 moves to VNCP2, *and* VNCP1 is blank. 

If I migrate both resources back to node N1, VNCP2 goes empty and the volumes show up 
in VNCP1. 

That is not the behavior I have been told existed on NetWare. Is that how it works in OES2? 

In any event it does not seem like correct behavior. It appears that only a single bind can 
exert influence on any given node, any other binds are either ignored, or bungle things. 

So, the idea that there is one VNCP object per "resource" does not seem to hold up. 

End Of Transmission. 

joe a. 

>>> "joea at j4computers.com" <joea at j4computers.com> 05/15/09 10:02 AM >>> 
This is an an attempt to backup and regroup. If we get nowhere with this 
go round, I will attempt to remain silent on the topic. 

If I understand things, and I believe I do, I have had this functional, since, 
almost, the beginning. 

When I bind the VNCP server to the "resource" IP, whatever volumes are 
mounted show up under the VNCP object. 

The problem appears to be that, once a resource IP is bound, I see all the 
mounted volumes. Not speaking of the SYS and _ADMIN volumes, but of 
any volumes other resources may have mounted on that node. 

For example, I have a two node cluster. Resources R1 and R2 are loaded 
on node "N1". Each resource has a volume created and mounted at load 
time, V1 and V2. When I bind either resource IP to the VNCP name, both 
volumes, V1 and V2, appear. 

Now, if I dismount V1, V1 disappears from the view of VNCP. If I mount V1, 
V1 reappears in VNCP. 

If I migrate R1 to N2, V1 also disappears from VNCP. As one might expect. 

However, if I then attempt to bind R2 to VNCP, the bind completes without 
error report, but, V1 never reappears in VNCP. 

Is that normal behaviour? 

If so, that seems a serious limitation, for those who have more than one 
resource on a cluster. 

If not, goodness, I am doing something wrong. Or, Horrors, there is a bug. 

joe a. 



_______________________________________________ 
Novell mailing list 
Novell at netlab1.oucs.ox.ac.uk 
http://netlab1.usu.edu/mailman/listinfo/novell

_______________________________________________
Novell mailing list
Novell at netlab1.oucs.ox.ac.uk
http://netlab1.usu.edu/mailman/listinfo/novell