Major version change: convert vs. D&L

[Rob Fitzpatrick]

For those of you who have been through a few major version changes, I'm interested to know if you have a preference for how you move your databases to the next version.

Do you just run the proutil conv(n)(n+1), or do you dump and load into a DB created with the new version?

Do you treat all DBs the same (prod/QA/dev/test/etc.), or do you do something different for production DBs?

Are there "gotchas" you've experienced in the past that I should be wary of? (Aside from the usual due caution; taking backups and so forth.)

Based on some brief testing it appears I do not have to run proutil updateschema and proutil updatevst after a conversion. Correct?

 

[TomBascom]

I think you only need to run the upd* options if the conversion spans versions where there are changes.

Every case is different but I tend to do a conversion, recompile, let things settle and then do a dump & load to make whatever changes are needed to leverage the new stuff. I likewise wait a little while before introducing new parameters.

That way any code and language types of issues are more clearly separated from any issues related to database features and capabilities.

 

[Rob Fitzpatrick]

Thanks Tom. My thoughts were specifically around moving from 10.2B to 11.x. There are some new schema tables and some DB changes.

 

[cj_brandt]

we run the conv910 script to upgrade db. Then at a point down the road, a dump and load is performed to take advantage of new features.
8 to 9 we implemented new Storage Areas.
9 to 10 was Type II SA.