I've been using our test setup for further testing, and I'm thinking of 
reversing my strategy. I may want to upgrade clients first, and then servers.

The basic issue is still how to overcome the roadblock of having an 
Administrator ID automatically switched from TRANSITIONAL to STRICT upon first 
login from a 7.1.8/8.1.2+ dsmadmc client. IBM seems to think we can upgrade all 
servers and all clients to 7.1.8/8.1.2+ simultaneously. That is not practical.

In the worst case, this automatic switching could cause the System 
Administrator's worst nightmare - to lose control over a running system. 

I am still considering the (very ugly) bypass of an administrative schedule 
that sets it back to TRANSITIONAL for all Admin IDs every 5 minutes. There will 
still be some failures.

But I am also considering reversing the strategy I had considered earlier, to a 
different strategy of upgrading all of the clients involved (about 7 of them, I 
think, but I'm not sure) to 7.1.8 or 8.1.4 first, while the servers are all 
still running older versions. So far, everything would be working.

Then doublecheck that there are not any left behind by scanning activity logs, 
the summary file, etc. 

Then once the operation of these clients was stabilized, upgrade our 4 servers 
one at a time. As each server is upgraded, the already-updated client would 
cause certificates to be exchanged and that Admin ID to be switched to STRICT, 
which would be OK since all of the client nodes where that Admin ID might log 
in from would already be at V7.1.8/8.1.2+. (At least we hope. This may expose 
those we forgot.)

Unless I'm overlooking something big here, I think this would allow us to 
upgrade each client and each server independently, and iron out any issues one 
at a time. Any comments on this client-first strategy?

Roger Deschner
University of Illinois at Chicago
"I have not lost my mind; it is backed up on tape somewhere."