Yeah, that's a good question. We don't have TCP keep alive enabled (default
out-of-the-box setting in the core has them off too), that's something we could
try out as well, but, in my past experience tcp keep alives have caused other
weirdness. Does anyone have positive experience and recommend turning them on?
But, I think independent of the bug, the overridable part still makes sense for
the other reasons (e.g consistent with the max conns setting ;) )
Thanks,
Sudheer
On Friday, August 16, 2019, 12:27:14 PM PDT, Alan Carroll
<solidwallofc...@verizonmedia.com.INVALID> wrote:
Is ATS actually receiving FINs on those connections? Can ATS know the
session is dead before trying a transaction on it?
On Thu, Aug 15, 2019 at 7:45 PM Sudheer Vinukonda
<sudheervinuko...@yahoo.com.invalid> wrote:
> Use case - We’ve some origins behind our proxy that get very low qps. Our
> server session sharing is set to Host based and have a floor on min keep
> alive connections as it is not overridable at the moment. Because of the
> floor the server sessions never drop below that number, even when those
> individual servers are down or taken out of rotation. When a request
> finally arrives for these services, because of host based match, we tend to
> pick up these stale sessions, ignoring the new dns results. This can go on
> for a while failing the requests. This is actually a bug in that we aren’t
> purging the stale sessions (bcoz of the floor handling) from the session
> pools.
>
>
> So, we’d like to customise the floor min keep alive connections per
> service. Here’s the PR https://github.com/apache/trafficserver/pull/5830
>
> Any comments/suggestions?
>
> Thanks,
>
> Sudheer
