Re: Compare libraries across systems or on the same system

On 06-Jan-2012 07:56 , Charles Wilt wrote:

Trying to compare libraries, ie. the complete set of objects in a
particular library vs. the set of objects in another library, either
on the same machine or on a different machine.

Right now, I'm doing so via querying the output of DSPOBJD like
so... with <<snipped code adding prefix to columns>>
, FO
as ( select * from SYSA left outer join SYSB
using (ODOBNM, ODOBTP, ODOBAT)
UNION ALL
select * from SYSA right exception join SYSB
using (ODOBNM, ODOBTP, ODOBAT)
)
select * from FO
where o_odlbnm is null or p_odlbnm is null
or o_odobsz <> p_odobsz or o_odcpvr <> p_odcpvr
or o_ODSRCD <> p_ODSRCD;

All boxes are at v5r4, and should be at the same PTF level...

First of all, is there a better way to do this?

I have written similar comparisons at least a dozen times, each time with modified considerations. I never concluded any one was better than another, because each was specific to the needs at the time. I seem to recall someone saying "management central" had some capability to compare across systems, but I never used that product so can not confirm such a feature.

Second, I notice that some of the differences in sizes don't seem to
have any reason...

Example for one PF, which has 0 members and the same format on box boxes:
System A, object size 20480
System B, object size 28672

Likely the shared *FMT object [possibly other shared objects too], as described in another reply; see DSPDBR RCDFMT(*ALL) for both files.

I think I recall reading something about why this is the case,
perhaps because system B has 6TB of disk and System A has only
0.85TB?

I am not sure of available storage, but the page size could matter, in that the minimal object size due to default page size.?

Is there anyway to account for this in my comparisons?

Since the *MEM object is a sub-component of the database *FILE each with their own "sizes", those objects might best be compared versus comparing their *FILE as container.

The sizes [dataspace and index] is still problematic for comparison, because a member with identical data could still mis-compare; a rebuilt index may size differently than one that was maintained over time. Will a difference in size be generally relevant when the data is identical or at least where the data-sets match?

Regards, Chuck