A default invocation of dbvalid will execute a VALIDATE DATABASE as well as a VALIDATE TABLE for each table. From what I see in the v11 code, validating a table will also validate each index on the table. In v12, things are implemented very differently and it's more difficult for me to determine off-hand exactly what is being done. I can tell you that we erroneously left the EXPRESS CHECK syntax and options in v12 but they are ignored by the server since they are no longer needed. We do something much cooler instead. In short, yes, it would appear the dbvalid default is to do everything.

@John: Thanks for the details! - But the mere fact you did check code of different versions seems to second Breck's point that the docs are not clear here. And don't get me wrong - I do not want to check the code myself. But I would prefer to know what kind of validation I should practise (based on different risk assumptions). In that respect, your statement that the V12 dbvalid default does "everything" (or "everything senseful"?) is very welcome.

The documentation for dbvalid is lacking because it just uses a wordy description rather than either describing which SQL statements are executed or the exact behaviour directly. The description of the SQL statements isn't too bad but could be improved. For example, VALIDATE TABLE says that it does checksum verifications so that would mean that an extra checksum validation wouldn't be necessary; however, there's no real way to get to that info from the documentation of the dbvalid tool. Even if the doc were fine, I would have gone to the source to verify.

So far, btw, I've only said that v12 is different and its exact semantics are harder to look up so I haven't done that yet. Still, the intention is that a default run of dbvalid should be sufficiently extensive test of database validity. Other options are largely for doing a subset of tests (only some tables, only some indexes, just page checksums, etc).

VALIDATE DATABASE visits every page and performs some other sanity checks. By virtue of reading each page, checksums are verified. So, VALIDATE DATABASE is effectively a superset of VALIDATE CHECKSUM. FWIW, VALIDATE DATABASE does its work by hauling everything through the cache but VALIDATE CHECKSUM reads directly from the file. I'm not sure why -- not my code :)

What does it mean to "check data correctness"? That could mean many different things. If checksums are enabled, VALIDATE DATABASE does verify that data hasn't changed while it was out on disk. It doesn't verify that the page wasn't

[I wish comments could be longer]...scribbled on prior to be written out last (in which case the checksum is computed after the corruption occurred). VALIDATE DATABASE also does not verify that the set of values indexed in each index matches the set of values in the base table. VALIDATE TABLE for each table accomplishes that last part.

@John: Thanks again for taking the time to explore this issue - that answers my (and Breck's) questions.