[SUMMARY: Deleting a member node from TruCluster]



Dear Managers,

My original questions were as follow.

I have a two-node TruCluster using Tru64 UNIX version 5.1A. I need
to remove (delete) the member upon which the cluster software was
first installed and the cluster first built (I refer to it as the
master node). Now I remember reading some documentation a long time
ago that any addition or deletion of members of the cluster should be
done from the master node (using clu_add_member or clu_delete_member
as the case may be).

Now my problem is what to do if I need to remove the master node?

Any advice, caveats or war stories will be highly appreciated.

Can anyone rember which was the document that would have mentioned
that master node?


With thanks to Dr. Thomas Blinn, Raul Sossa and Dominic Christopher, my
problem is solved.

It is possible to delete any non-running member of the cluster from any
running cluster member without any problems. In particular, I will
quote for you the reply from Dr. Blinn, as his reply was a detailed
explanation.

The "master" is the system on which Tru64 UNIX and TruCluster software were
originally installed and upon which the "clu_create" command was issued to
build the cluster. That system was originally running as member 0 in the
context of what became the cluster. It created a cluster bootable system
from the original installation, and by default would have chosen member 1 as
the identifier for its own cluster member specific files. Later, if you add
a new member with "clu_add_member" a different member number is chosen, so
that each member has a unique member number.

Because the cluster is "symmetric", you can use any running member to remove
any non-running member's part of the file system from the cluster. There is
not any particular special role for any particular member. The one thing
you can't readily do is wipe the boot disk of a member that has a private
boot disk; that is, if member "a" can't access member "b"'s boot disk and
you use member "a" to remove member "b" from the cluster, it won't be able
to wipe member "b"'s boot disk.

This stuff is all documented in the cluster admin manual, as I recall.

Tom

Additionally, Dominic Christopher reported actual experience successfully removing and adding back the "master" member node without any problems.

Thank you all.

Yours sincerely,
Robert Honore.



Relevant Pages

  • Deleting a member node from TruCluster
    ... I need to remove the member upon which the cluster software was first installed and the cluster first built (I refer to it as the master node). ... Now I remember reading some documentation a long time ago that any addition or deletion of members of the cluster should be done from the master node. ...
    (Tru64-UNIX-Managers)
  • Tar Restore problem
    ... I took backup using tar on tru64 5.1A & when I tried to restore on another ... So I rebuilt cluster_root exactly like documented in the "cluster ... The same is true if I boot the first member (without the kernel clubase:.. ...
    (Tru64-UNIX-Managers)
  • SUMMARY: Cluster crash
    ... nature to save only filesystems. ... Betreff: Cluster crash ... make it new but on the same disk and with a fresh backup restore. ... The same is true if I boot the first member (without the kernel clubase:.. ...
    (Tru64-UNIX-Managers)
  • Process are blocked for several seconds in a cluster
    ... I have an application running on one member as a cluster service. ... This application communicates over TCP/IP with a remote host and ...
    (Tru64-UNIX-Managers)
  • Re: Multiple Clusters with shared passive nodes
    ... Sort of - I'd create all 5 servers as members of a single cluster, then set up 4 instances of SQL Server with cluster server member 1 configured ... While you are able to have cluster member 5 be the failover for all the other instances, there might be a better way to set which member servers can run each instance, but its not difficult to change this at a later date - it might not be an optimal configuration if all the instances are of different memory needs or if you want to handle a situation where more than one cluster member might be unavailable at a time. ...
    (microsoft.public.sqlserver.clustering)