Tom (and others who responded),

The "feature" of zero-down time and the demonstration at WWRUG 2012 was a lab 
demo to show a
concept that we had taken from design into prototype to gage interest from the 
customer base.

To be clear, it is not present in any release of the product at this time.

We are still gaging customer interest and I will say unfortunately, the 
interest expressed in this topic has
not been as strong as several of us expected.  There is interest – don't get me 
wrong – but there are other
things that are generating much more concern from customers.  There is still a 
contingent struggling to get
this capability added to an upcoming release – but nothing has made it yet.


The functionality we demonstrated was true zero-down time.  It creates parallel 
metadata tables with new
and old definitions and different servers pointing to different sets – ALL ON 
THE SAME DATA TABLES.  Lots of
work went in to handle deleted fields (delete is deferred because the old 
version is still using the fields) and
to even handle cases where archgid changes field IDs (extra view forms are put 
in place so that both old and
new use the IDs they expect and all is in place).

Both old and new versions of the applications were available and running 
through different servers in the
server group so you could confirm that all was ready before you flipped the 
switch.

If any system went down, it came back up looking at its set of definitions so 
everything is robust and
recoverable while in this mixed mode.

We went so far as to have servers running the old app have a signal to load and 
prepare the new definitions
in the background while continuing to run the old version so when they were 
signaled to go live, they finished
any current API call in the old version and any new API call started working 
with the new version for absolute
zero down time transition at the server.

This means that any program or integration has zero interruption or disruption.

Now, the interactive user through the mid-tier has the issue of caching.  
Depending on whether an immediate
update occurs or you are just going to let the next interval check do it, it 
may be instant to a few minutes
before the updates are reflected to the client and there is going to be a 
performance slowdown that will be
noticeable as things are reloaded (kind of like a first time reload after 
mid-tier startup) and then changed
screens will show up.  So, they are never down, but there is a short time of 
affect on interaction of users.


This is not fantasy as we have prototyped it and it was what was demonstrated 
at WWRUG and it does work.

If you have not entered your vote for this feature on the BMC Communities AR 
System thread Ideas entry
for this feature (this is an area where you or BMC posts enhancement requests 
and the community can vote
on it to show interest – higher vote totals help push features up the list for 
implementation), please enter
your vote to show support for the idea.


This does not mean no work for the Administrator during an upgrade, but it 
means no end user outage for
an upgrade.  It should also help reliability and success of an upgrade in a 
dramatic way as you can test and
verify and make any corrections in the new version before go live.  It also 
should remove the I have to do
everything under time pressure in the middle of the night on a weekend issue to 
avoid customer downtime
as much as possible topic.  So, a win for everyone all around.


I hope this confirms and clarifies this capability, what it is and where it is.

Doug Mueller

From: Action Request System discussion list(ARSList) 
[mailto:arslist@ARSLIST.ORG] On Behalf Of Tom Shurmur
Sent: Monday, July 15, 2013 6:57 PM
To: arslist@ARSLIST.ORG
Subject: 7.6.04 Upgrade to 8.1

**
Howdy Listers,

We are about to embark on upgrading from 7.6.04 to 8.1 on a windows platform 
using MS-SQL. Has anyone used the No Downtime Upgrade method that was presented 
at last year’s WWRUG to move from 7.6.04 to 8.1? If so, was it smooth or bumpy 
ride?

We will be standing up a test system with 2 app VMs and a DB to test this 
approach.

I look forward to your feedback

Tom Shurmur
Sr Remedy Developer
Froedtert Health System
_ARSlist: "Where the Answers Are" and have been for 20 years_

_______________________________________________________________________________
UNSUBSCRIBE or access ARSlist Archives at www.arslist.org
"Where the Answers Are, and have been for 20 years"

Reply via email to