On Tue, Sep 03 2013, Grant Grundler wrote: > On Sun, Sep 1, 2013 at 7:39 PM, Grant Grundler <grundler@xxxxxxxxxxxx> wrote: > ... > >> Sure. "data_rentention_check" would better as a chapter title for > >> documentation. I'll see if the existing "--readonly" flag could > >> behave like I expect. > ... > > After talking with Juan, it seems --readonly isn't a good option. One > problem is the current implementation does parameter checking (which > is correct). But what I want is to in fact not do any "workload" IO > at all. > > We will preserve the existing behavior (specifying "read only > workload") will result in verify phase checking all the fields that > were checked before - it will NOT check numberio. Using numberio to > detect stale data is a new feature which breaks the existing behavior > and I don't want to be responsible for fixing something I don't like > in the first place. > > However, to avoid re-running the entire (readonly) workload again > just to verify the data, I'm proposing "--verify_only" and > "verify_only=bool" flag to indicate: > 1) implies do_verify=true and verify=meta > 2) FIO will run through IO workload steps > 3) not do any IO during the workload phase > 4) verify phase will run ONCE after all iterations of the workload > have been "dry-run" (when we know what numberio should be for each > block). > > The goal is to re-use existing code to generate the state of numberio > for each block to what we expect at the end of any workload that > performs writes. > > Thoughts on "verify_only=bool" flag? This suggestion makes more sense to me. I'd be fine with a verify_only bool option, THAT makes sense and clearly states what it does. -- Jens Axboe -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
