Conference smurf::ase

1975.0. "Unable to remove a member." by MEOC02::JANKOWSKI () Tue Apr 01 1997 06:39

    Hi,
    
    I am unable to remove a member from DECsafe configuration.
    As I am unable to remove it I cannot configure it back.
    What can I do?
    
    The full story follows:
    
    I have a 2 member configuration - members are named hx11 and hx12.
    There is only one service - sapdb - SAP R/3.
    Current configuration is DU V3.2G, DECsafe V1.3, 2x 8400 with 4GB memory
    and 4x440MHz processors each. There are no patches installed.
    
    The site has a long and painful history of DECsafe.
    After a 10 months struggle the configuration is still not working.
    
    At some point in time the member hx12 was disabled - ASE="off"
    in /etc/rc.config due to some operational problems.
    So for a long period of time the configuration was down to one member (hx11).
    Effectively DECsafe was used to boot the service.
    In the meantime hardware, operating system, and application have been
    upgraded several times.
    
    Yesterday we were trying to activate the dormant member - (hx12).
    The following happened.
    
    1.
    We set ASE=on on hx12 and rebooted. hx11 was up at the time.
    The system (hx12) panicked during reboot immediately after ASEAM got
    started.
    
    2.
    We then reset the ASE configuration on hx12 to initial state using
    standard DECsafe provided utility.
    
    3.
    We then tried to remove hx12 from asemgr running on hx12 whilst hx11
    was shut down.  This did not work. We just got heaps of messages every
    5 seconds about inability to do the removal.
    
    4.
    We then brought up hx12 and tried to delete hx12 from asemgr running
    on hx11. The result was that this time hx11 panicked.
    
    We recovered again and ran out of options.
    
    Our objective is to bring hx12 back as an active member to the working
    configuration without deleting the existing service from hx11
    and reinstalling everything from scratch.
    This machine runs heaps of disks in a complex HSZ/LSM/AdvFS configuration.
    Adding them again from scratch is risky and time consuming.
    We cannot afford it due to the risk x time factor.
    
    We have a working configuration.
    We just want to delete the inactive member and then add it again.
    What can we do to achieve our objective?
    
    Regards,
    
    Chris Jankowski
    Melbourne Australia
    

Chris,

It looks like things could have gone in a number of different places
during your description.  It should be noted that some of the things
done are most likely not supported (ie. upgrading members out of sync,
initializing a member before deleting it, etc).  With that said, lets
look for a solution to solve your problem.

My suggestion is to get hx12 back to a known state.  Try the following.
On hx12 do:

	rcmgr set ASE on
	rm /var/ase/config/asecdb
	/usr/sbin/asesetup
	reboot (being this drastic is not necessary, but removes variables)
	ps ax | grep ase (make sure ASE daemons are running, aseagent, etc).
	Make sure message "Agent Notice: in install state" is in daemon.log

On hx11 try to again remove hx12 from the ASE.

Verify that all sofware and hardware on both systems are at the same revs.

Then re-add hx12 back into the ASE from hx11.


I tested this procedure on my test systems running a newer version and it
worked.  The same thing should work in V1.3.  If this does not work for you
I would suggest a QAR to get someone to look a little closer at your
environment.

-- Greg
[Posted by WWW Notes gateway]

    Hi,
    
    We applied the DECsafe V1.3 patches and re-initialize the DECsafe
    database as in .1. However, we encountered another problem described
    in note 2068. Any help appreciated...thanks!
    
    K. Lee
    NSIS Melbourne