Re: BRMS - full save with least amount of restricted time

[Jim Oberholtzer <midrangel@xxxxxxxxxxxxxxxxx>]

Evan,

I can't speak to your experiences but I have done dozens of recovery 
tests and I've only run into trouble with QUSRSYS or QGPL on very 
limited cases, and not affecting TCP.

*SYNCLIB is a beast in that you have to get all the libraries in the 
list to checkpoint at the same time.  Important for some folks but if 
you do not have user database objects (meaning non-IBM) in either 
QUSRSYS or QGPL, I would not include them in the synch point.  Saving 
QUSRSYS and QGPL with a synch point of *LIB has always been successful 
in the tests I have run.  Again there may be some anomalies but as a 
general rule the save works quite well.

A better reason to test your recovery I can't come up with.  If you have 
not tested your recovery, you don't have a reliable back up, regardless 
of what the system says it put on tape, virtual or physical.

Jim Oberholtzer
Chief Technical Architect
Agile Technology Architects


On 1/22/2013 12:24 PM, Evan Harris wrote:
> Hi JIm
>
> I recall being unable to save QUSRSYS using SWA on V5R4 while TCP was
> active. More accurately I seem to remember that some items just would
> not restore correctly even though the save seemed to complete OK.
> These items seemed to be associated with TCP jobs. The SWA was a case
> of quiescing user applications until the checkpoint (*SYNCLIB) was
> reached, except for downing TCP. To get around it I modified the save
> regime to save QUSRSYS weekly or monthly when TCP was down.
>
>
> On Tue, Jan 22, 2013 at 10:09 AM, Jim Oberholtzer
> <midrangel@xxxxxxxxxxxxxxxxx>  wrote:
> > >  Jeff answered most of your questions, I'll deal with the QUSRSYS and
> > >  recoverability.
> > >
> > >  Yes the system is completely recoverable with the SWA on QUSRSYS, in
> > >  fact I have performed many recovery tests using BRMS, and in many ways
> > >  it works better than the manual recovery.  Were you using BRMS for the
> > >  SWA processing?  If not I suspect that the checkpoint processing my not
> > >  have been done quite right leaving some objects out of synchronization.
> > >
> > >  It is also true that a slowdown in normal activity will make the SWA
> > >  work much better as well.  The best situation is to be able to quiesce
> > >  the the system for a short time (about 5 minutes in my experience) while
> > >  the checkpoints are taken, particularly with *SYNCLIB or *SYSDFN check
> > >  points.
> > >
> > >  Jim Oberholtzer
> > >  Chief Technical Architect
> > >  Agile Technology Architects
> > >