Runstats / OLTP environment

[Michel Esber]

Environment: DB2 v8 LUW FP 15 Linux.

I have an OLTP system and several applications run frequent and
concurrent insert/delete/select statements.

Table sizes can grow a lot, and eventually I need to collect new
statistics.

While trying to use 'allow write access', this following warning is
returned:

SQL2314W Some statistics are in an inconsistent state. The newly
collected "TABLE" statistics are inconsistent with the existing
"INDEX" statistics.

Using allow read acces is not an option, unfortunately. I´ve tried
table and index sampling, but that also did not help much.

My question is: how bad can this 'inconsistency' be ? Can I safely
ignore it ? Or can this lead my system to poor query planning and
statement reponse times be slower than if statistics where taken using
'allow read access ?

Thanks, Michel.

[Mark A]

On Aug 23, 10:07*pm, Michel Esber <[e-mail]:

Environment: DB2 v8 LUW FP 15 Linux.

I have an OLTP system and several applications run frequent and
concurrent insert/delete/select statements.

Table sizes can grow a lot, and eventually I need to collect new
statistics.

While trying to use 'allow write access', this following warning is
returned:

SQL2314W *Some statistics are in an inconsistent state. *The newly
collected *"TABLE" statistics are inconsistent with the existing
"INDEX" *statistics.

Using allow read acces is not an option, unfortunately. I´ve tried
table and index sampling, but that also did not help much.

My question is: how bad can this 'inconsistency' be ? Can I safely
ignore it ? Or can this lead my system to poor query planning and
statement reponse times be slower than if statistics where taken using
'allow read access ?

Thanks, Michel.

All it means is that the number of rows in index (when examined by
runstats) is different than the number of rows in the table (when
runstats looked at the table immediately prior to that). This happens
becasue their are deletes and inserts happening while the runstats is
running. The differences are not likely to be significant, so just
ignore the warning.