RE: PowerHA for i questions

<snip>
if the routers have the capability
</snip>

We JUST upgraded to such routers.

We've long been a Mimix shop.  Which is different than PowerHA (the title 
of this thread).
We have 
U78C9.001.WZS00UA-P1-C7  10 Gigabit Ethernet-SFP+ SR PCI-E adapter
assigned to one vios lpar and
U78C9.001.WZS00UA-P1-C6  10 Gigabit Ethernet-SFP+ SR PCI-E adapter
assigned to another vios lpar.  Both are utilized by all guested lpars. 
Both serve up four vlans.  This is for redundancy. We have some guests who 
are in our DMZ with external IP addresses.  They use these same adapters. 
We have some guests that are simple internal use only and they use these 
same adapters.  We have some guests that are mimix lpars and they use 
these same adapters.

A Mimix lpar may look like:
wrkhdwrsc *cmn
CMB08           268C  Operational           Comm Processor
  LIN04         268C  Operational           LAN Adapter 
    CMN04       268C  Operational           Ethernet Port 
CMB09           268C  Operational           Comm Processor
  LIN03         268C  Operational           LAN Adapter 
    CMN03       268C  Operational           Ethernet Port 

wrklind
Line        Type        Text 
LANLINGB    *ELAN       GB Ethernet line 
LANLINMX    *ELAN       Mimix: GB Ethernet line

netstat *ifc
Internet         Network             Line      Interface
Address          Address          Description  Status 
10.10.6.32       10.10.6.0        LANLINGB     Active 
10.10.6.33       10.10.6.0        LANLINGB     Inactive 
10.10.252.192    10.10.252.0      LANLINMX     Active 

This is a backup mimix lpar, GDIHQ2.  The primary is GDIHQ1.  The users 
only know them as GDIHQ.  All mimix replication traffic goes through the 
'252' LANLINMX interface

The primary looks like this
Internet         Network             Line      Interface
Address          Address          Description  Status 
10.17.6.32       10.17.6.0        LANLINGB     Active 
10.17.6.33       10.17.6.0        LANLINGB     Active 
10.17.252.190    10.17.252.0      LANLINMX     Active 

Notice that the primary is in our '17' subnet in one city and the backup 
is in our '10' subnet in another.
When we do a switch the users have a brief outage.  In that outage the DNS 
is changed to say GDIHQ is 10.10.6.33 instead of 10.17.6.33 and both 
10.17.6.* addresses are stopped.  The 10.10.6.33 is made active.

Tricks are things like making sure the relational database directory 
entries (WRKRDBDIRE) are ok so that 'users' can go to GDIHQ all the time 
and systems can get to GDIHQ1 or GDIHQ2 if they need to query a particular 
machine.  No problem.

GDIHQ1:
                    Remote 
Entry               Location
GDIHQ               *LOCAL 
GDIHQ1              GDIHQ1 
and if you look at the second one you will see:
Relational database  . . . . . . :   GDIHQ 
  Relational database alias  . . :   GDIHQ1
Remote location: 
  Remote location  . . . . . . . :   GDIHQ1

GDIHQ2:
                    Remote 
Entry               Location
GDIHQ               *LOCAL 
GDIHQ2              GDIHQ2 
and if you look at the second one you will see:
Relational database  . . . . . . :   GDIHQ 
  Relational database alias  . . :   GDIHQ2
Remote location: 
  Remote location  . . . . . . . :   GDIHQ2

We also have BRMS network and advanced.  BRMS doesn't care about GDIHQ; 
just GDIHQ1 and GDIHQ2.
We back up on GDIHQ2 daily.  We only backup on GDIHQ1 quarterly.  We do 
not even keep tapes in the city where GDIHQ1 is located until the time of 
that quarterly backup.

We do switches every quarter.  Granted, we only run on the backup machine 
during that downtime weekend.  But the users can connect and work. 
Including traditional 5250 and ODBC.  They got their spool files and all.

When we do an OS upgrade we upgrade GDIHQ2 during the week.  That Friday 
we do the switch and we're running on GDIHQ2.  We upgrade GDIHQ1.  We 
switch back.

Keep in mind that a switch has three phases:
P1:  Switch from Primary to secondary
P2:  You're done doing upgrades, full system saves, etc on Primary.  Start 
data resync.
P3:  Switch back users from backup to primary.

We keep about two years of history of these switches, and all the details 
of what was done (including times) in a file cabinet.



Rob Berendt