Hi

I was going to just add this to the previous email about logging, but
thought it deserved its own as it may be of use to someone about to
migrate.


Keeping listeners going on both endpoints has been quite successful.  
For one thing it enabled me to easily see who was still coming in on the
wrong IdP.  I used the technique suggested to me by Adrian Barker: 
Idp.dundee still comes into the same Apache and that box still has its
Shib1 IdP.  That Apache has a proxypass to the box that hosts Shib2.  
Probably today I'm going to get the DNS switched over to have all the
traffic now go to the Shib 2 box prior to flattening the old Shib 1 IdP
and upgrading that too.

Edina updated the metadata at 18:15 on Thursday night.  By 07:30 on
Friday morning it seemed like the vast bulk of requests were coming into
the shib 2 IdP.
By midday only these SPs were still coming into Shib 1:

Adept Scientific (Adept4Education) (ADEPT) 
British Periodicals collections 1 & 2 (BRITPERIO) 
Early English Books Online (EEBO) 
House of Commons Parliamentary Papers (HCPARLP) 
John Johnson collection (JOHNJ) 
* JSTOR journal archive (JSTOR) 
Literature Online (LION) 
* SCOPUS (SCOPUS) - 
* SpringerLink journals (SPRINGER)
* Westlaw
heinonline
Web of knowledge

JISCMail had just transferred at 11AM but were actually broken until
then (as were the last two above).   There was quite a common issue that
caused a few to break - split personaility:   An SP would update
metadata in the authentication part, so when you went there you would go
to the Shib 2 IdP, but after authenticating you would see an attribute
request come in to the Shib 1 IdP!   (I think Hein were the other way
round).   Of course this ain't going to work.

By 16:00 on Friday I had this report from the library:

—----------
Shib 2 and not working
a-n The Artists Information Company - re-checked at 15.50pm still going
to Shib 2 address but still not working once logged in 

Still Shib 1.3 and working

Adept Scientific (Adept4Education) (ADEPT)  checked morn 11/12/09
Re-checked 15.32 - same

SCOPUS (SCOPUS) - thinks on campus - still thinks we are on campus
re-checked 15.43 

SpringerLink journals (SPRINGER) - re-checked 15.44 still Shib 1.3

Westlaw  - re-checked 15.46 still Shib 1.3

The following Shib 1.3 from this morning are now working from Shib 2

British Periodicals collections 1 & 2 (BRITPERIO) checked morn
11/12/09
Re-checked 15.33 - moved over to Shibb 2 (Proquest)

Early English Books Online (EEBO)  checked morn 11/12/09
Re-checked 15.35 - moved over to Shibb 2 (Proquest)

House of Commons Parliamentary Papers (HCPARLP) checked morn 11/12/09
Re-checked 15.38 - moved over to Shibb 2 (Proquest)

John Johnson collection (JOHNJ) checked morn 11/12/09
Re-checked 15.39 - moved over to Shibb 2 (Proquest)

JSTOR journal archive (JSTOR) checked morn 11/12/09
Re-checked 15.40 - moved over to Shibb 2 (Proquest)

Literature Online (LION) checked morn 11/12/09
Re-checked 15.41 - moved over to Shibb 2 (Proquest)
—----------

So you can see there was a gradual transition over the day.   One point
I should make is that Westlaw are flagged as working - this is only
because they were doubly broken!  We have a wayfless session initiator
URL which still  went to the shib 1 IdP for authentication and got
attributes from there too- fine(ish).  But if you went in through the
wayf it authenticated on shib 2 but still asked for shib 1 attributes -
bust!   This has just been fixed last night by a server reboot their end
(which they could only do now).    Now all links for westlaw go through
Shib 2.   

Heinonline were quick to reply to my email Friday with a fix as were
Adept Scientific when I emailed them (yesteday I think),  both these
were in response to emails to the addresses in the metadata.     Emails
to those addresses for Web of Knowledge and Westlaw have proved less
productive and we've had better results going through the UK sup
port
contacts.

At start of play Wednesday As far as I can tell only Web of Knowledge
are still broken - the library have put in place IP based access for on
campus users I think, for the interim.   Unfortunately for WoK that now
means I have no other distractions and they have my full focus ;-)......
   I suspect though there's little we can do - the folks at Mimas are
working hard on our behalf, but it's Thomson Reuters who have to do the
fix.

My conclusions:

The migration was mostly harmless
There was a little disruption but there were workarounds that the
library used.
Keeping listening for Shib 1 and Shib 2 was a good ploy.


If there's anything else of use I can add please ask.

HTH someone who's going down this (frightening) path ;-)
Andy



The University of Dundee is a registered Scottish charity, No: SC015096