Only thing I'm worried with on this is that repair tends to be slower for me than upgradesstables in practice whenever I've tried to do something similar. And you still will need to do a cleanup afterwards anyways.
Sometimes just leaving it alone and not poking every two seconds might be the best bet (assuming C* isn't dying while your trying to do the upgradesstables op)
If you are reading at quorum you are ok. However if your reading at one you suffer a chance of getting an empty return until repair is complete.
On Thursday, February 28, 2013, Hiller, Dean <Dean.Hiller@nrel.gov> wrote:
> My goal is tweaking the bloomfilter down and needing to run
> upgradesstables is really not working out and I figure wipe the CF on one
> node and then run repair to get all his data back with the new
> bloomfilter. We backed up one node and tried this and it seems to workİis
> this ok or will I have issues?
> Why is it not working out you askİ.Well, we keep seeing an upgradesstables
> on our nreldata CF go to 100% and then start all over again at 0% so we
> killed it as it is already taking over 24 hours not sure it is getting
> On 2/28/13 12:25 PM, "Michael Kjellman" <firstname.lastname@example.org> wrote:
>>Is your goal to rebuild a cf from scratch?
>>On Feb 28, 2013, at 11:24 AM, "Hiller, Dean" <Dean.Hiller@nrel.gov> wrote:
>>> Title says it all. Anyone know?
>>Copy, by Barracuda, helps you store, protect, and share all your amazing
>>things. Start today: www.copy.com.
Copy, by Barracuda, helps you store, protect, and share all your amazing
things. Start today: www.copy.com