If I can reuse cache entries creating during failover, it means I can absorb failures without the cache hit rate suffering.
On Tue, Jan 5, 2016 at 8:20 AM, Oleg Kalnichevski <ol...@apache.org> wrote: > On Tue, 2016-01-05 at 07:30 -0500, Sam Perman wrote: > > If the url changes from one request to the next without a special > client, wouldn't the responses be two separate cache entries? > > > > Why is that bad? > > Oleg > > > sam > > > > > On Jan 5, 2016, at 7:00 AM, Oleg Kalnichevski <ol...@apache.org> > wrote: > > > > > >> On Tue, 2016-01-05 at 06:28 -0500, Sam Perman wrote: > > >> Maybe there is another way to implement what I'm trying to do? In my > case, if the backend request fails, I want to retry the request against a > different url, but still cache the response. > > > > > > Why do you need a special client for that? > > > > > > Oleg > > > > > >> thanks > > >> Sam > > >> > > >>>> On Jan 5, 2016, at 4:58 AM, Oleg Kalnichevski <ol...@apache.org> > wrote: > > >>>> > > >>>> On Mon, 2016-01-04 at 17:21 -0500, Sam Perman wrote: > > >>>> Hello > > >>>> > > >>>> I'm using the CachingHttpClient in a way that is deprecated in 4.5 > and am > > >>>> trying to figure out the right non-deprecated way to do this. > > >>>> > > >>>> In my cause, I have a custom implementation of HttpClient that will > do some > > >>>> request rewriting and retrying if it sees certain error conditions. > > >>>> > > >>>> This is what the old deprecated code looks like: > > >>>> > > >>>> CacheConfig config = buildCacheConfig(); > > >>>> HttpClient myBackendClient = new CustomHttpClient(); > > >>>> CachingHttpClient cachingHttpClient = new > > >>>> CachingHttpClient(myBackendClient, config); > > >>>> > > >>>> It looks like the recommended way to build caching http clients in > 4.5 is > > >>>> to use the CachingHttpClientBuilder, but I see no way to provide my > own > > >>>> implementation of HttpClient for use as the backend. Is this > possible to do > > >>>> in a non deprecated way? > > >>> > > >>> Sam, > > >>> You can't. The problem is that caching logic needs to be inserted at > a > > >>> particular point in the protocol processing chain in order to work > > >>> correctly in all cases. The approach of adding a caching layer as a > > >>> decorator for an arbitrary HttpClient instance turned out to be not > good > > >>> enough. > > >>> > > >>> Oleg > > >>> > > >>> > > >>> --------------------------------------------------------------------- > > >>> To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org > > >>> For additional commands, e-mail: httpclient-users-h...@hc.apache.org > > >> > > >> --------------------------------------------------------------------- > > >> To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org > > >> For additional commands, e-mail: httpclient-users-h...@hc.apache.org > > > > > > > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org > > > For additional commands, e-mail: httpclient-users-h...@hc.apache.org > > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org > > For additional commands, e-mail: httpclient-users-h...@hc.apache.org > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org > For additional commands, e-mail: httpclient-users-h...@hc.apache.org > >
