On Tue, Dec 15, 2009 at 11:29:33PM +0100, Krzysztof Ol?dzki wrote: > >>+error_limit <count> > > > >"error-limit" please :-) > > Sure. Would you like a new patch with that change?
No, don't waste your time on cosmetics. I've noticed you sent v3 while I was replying, I'll sed on v3. (...) > >Maybe I'm not very clear, please tell me if so. > > Yes, it is exacly how you described it - we'll be able to change > defaults without enabling the feature itself. Perfect then. > > Currently I'm thinking about something like: > >server default [error-limit XX] [on-error XX] [inter XX] [fastinter XX] > >[downinter XX] [rise XX] [fail XX] > + maybe some other options we might find useful. OK that would be cool. We'd just have to use a specific keyword such as "server-default" instead of "server default" in order not to create an ambiguous config setting (not creating a server called "default" and start to parse it). Hmmm please also add weight, minconn and maxconn that some people use a lot too ! (...) > This patch should not make healht checks more CPU hungry, however > failed requests cost now little more. Hopefully, if health analyses > feature is not enabled, it is one more call to health_adjust() which > simply returns in the first condition. OK that's fine. Once again, I was more bringing information from the field related to the area being touched than expressing concerns with this specific patch. > >Feel free to tell me when you'd like your patch to be merged. > >In my opinion it's already in good shape. > > I have just sent the v3 version, if you accept the changes I think we > can merge it now, of course with respect to s/error_limit/error-limit/g. OK I'm doing so then. Best regards, Willy
