Re: [PATCH] BUG/MINOR: lb: fix redispatch for hash based lb-algo's

2018-12-30 Thread Willy Tarreau
Hi Lukas, On Sun, Dec 30, 2018 at 09:52:33PM +0100, Lukas Tribus wrote: > In my tests your patch does the job and redispatch does work for > consistent source or uri hashing, but only if hash-balance-factor is > set. If I comment hash-balance-factor out, the redispatch does not > occur: > >

Re: [PATCH] BUG/MINOR: lb: fix redispatch for hash based lb-algo's

2018-12-30 Thread Lukas Tribus
Hi Willy, On Thu, 27 Dec 2018 at 15:54, Willy Tarreau wrote: > > I'm not 100% sure whether "option redispatch" was only intended to > > break cookie persistence, but not other lb algorithms. Docs are > > ambiguous about this. > > I think you meant hashing instead of cookie persistence. But

Re: [PATCH] BUG/MINOR: lb: fix redispatch for hash based lb-algo's

2018-12-27 Thread Willy Tarreau
Hi again Lukas, please have a look at this patch proposal. I think it could do the job, but I have only build-tested it, nothing more. Thanks, Willy diff --git a/include/proto/lb_chash.h b/include/proto/lb_chash.h index a0ebf696e..679dff363 100644 --- a/include/proto/lb_chash.h +++

Re: [PATCH] BUG/MINOR: lb: fix redispatch for hash based lb-algo's

2018-12-27 Thread Willy Tarreau
Hi Lukas, On Wed, Dec 26, 2018 at 11:07:27PM +0100, Lukas Tribus wrote: > > redispatch never worked for hash based alghoritms, as the code > > (BE_LB_LKUP_CHTREE -> chash_get_next_server()) would only have been > > called for BE_LB_KIND_RR, which doesn't make sense. Fix this by also > > going

Re: [PATCH] BUG/MINOR: lb: fix redispatch for hash based lb-algo's

2018-12-27 Thread Lukas Tribus
Hi, On Wed, 26 Dec 2018 at 23:04, Lukas Tribus wrote: > > redispatch never worked for hash based alghoritms, as the code For the commit: s/alghoritms/algorithms/ Lukas

Re: [PATCH] BUG/MINOR: lb: fix redispatch for hash based lb-algo's

2018-12-26 Thread Lukas Tribus
Hello, On Wed, 26 Dec 2018 at 23:04, Lukas Tribus wrote: > > redispatch never worked for hash based alghoritms, as the code > (BE_LB_LKUP_CHTREE -> chash_get_next_server()) would only have been > called for BE_LB_KIND_RR, which doesn't make sense. Fix this by also > going down this code path