[Info-ingres] Loading stats generates warning W_OP0973

[Martin Bowes]

Hi Stevo,

The short answer is yes. You're paying for service, you may as well use it.

But out of curiosity...

What z flags did you use in the optimisation call?

For example (using 11.1.0 +p15599 a64.lnx) on a table with 15000+ rows and an ingresdate field I did:
optimizedb -zk -zr1000 -zu1000 -zns -o stats.txt -uace ace_trove_live -rfield -atcounted

And then:
optimizedb -i stats.txt -uace ace_trove_live -rfield -atcounted

No errors, warnings, or informationals were generated.

You may like to experiment with the -zn# option as well. Try a -zn9 and see if that helps.

Martin Bowes

-----Original Message-----
From: Steve <s.anderson.au at gmail.com> 
Sent: 29 July 2021 04:16
To: info-ingres at lists.planetingres.org
Subject: [Info-ingres] Loading stats generates warning W_OP0973

Hi folks

Recently I generated tables stats in our DR environment, writing them to a file rather than the System Catalogs - that sounds awfully like the Ingres documentation :) - using the -o optimizedb flag. Later when loading the stats into our production environment, I received several W_OP0973 warnings.

Searching the Actian website, it seems the issue re warning W_OP0973 was fixed in 9.2 - below is the note from "Patch 14296 for Ingres 9.2 Service Pack 1 (9.2.1) on Sun Solaris UltraSPARC":

"Bug 122605 (GENERIC)

optimizedb generates incorrect exact histogram cell counts for date columns. Loading these stats via optimizedb will generate W_OP0973 warnings."

(NB: Am running Ingres 11.0 on Solaris SPARC)

Should I interpret these warnings to suggest the bug may not be fixed? Should I be raising this with Actian? 

Further, all but one of the columns with this warning are ingresdate columns. The exception is column defined as a float4.

Thanks
Steve
_______________________________________________
Info-ingres mailing list
Info-ingres at lists.planetingres.org
https://lists.planetingres.org/mailman/listinfo/info-ingres