Re: System i Navigator: SQL Plan Cache Snapshots; using Visual Explain for Index Advisor, to improve QZDASOINIT SQLperformance

Thanks, Chuck, for the clarifications - it's been far too long since I've played with the INI file.

Love the national language support on these boxes - it's automagic when you don't hard-code literals in your source - and this kind of thing, well, too cool to see the automagic conversion from Unicode.

Vern

On 3/9/2012 1:14 PM, CRPence wrote:

On 09-Mar-2012 08:51 , Vern Hamberg wrote:

If you do not have any select/omit logicals, you don't need to
change the derived-index item in QAQQINI. You want to look, though,
at the IGNORE_DERIVED_INDEX entry - all the entries are set to
*DEFAULT, what you should do is look at the description in the file.
Use RUNQRY () QAQQINI, DSPPFM doesn't show it, since the description
is in unicode, I think.

The job CCSID must be set to a value other than *HEX [aka 65535] to
enable the RUNQRY to effect any better output than the DSPPFM.

<<SNIP>>

Do NOT change the model file that is in QUSRSYS or in QSYS. Make a
copy of the one in QUSRSYS (I think - what you read will tell you)
and modify that.

The model file [with data] is shipped in QSYS. No file QAQQINI in
QUSRSYS should exist normally... and if one does exist in that library,
then the INI options from that file are effectively system values; i.e.
to avoid the global effects, CHGQRYA must be issued to change to use a
different QRYOPTLIB.

The definition of "make a copy" is specifically to issue CRTDUPOBJ
[or save\restore], and must include the triggers; i.e. CPYF or any other
data-only copy that creates the database *FILE named QAQQINI is not
acceptable for creating the INI options file. And although data is
optional, typically DATA(*YES) is desirable on the CRTDUPOBJ as well, to
enable UPDATE versus INSERT for most [i.e. those existing in the model]
INI options and settings.

<<SNIP>>

Regards, Chuck