https://issues.apache.org/bugzilla/show_bug.cgi?id=56119

--- Comment #30 from Sebb <[email protected]> ---
(In reply to Philippe Mouawad from comment #29)
> @Oliver,
> Really big thanks for your follow up and investigations on this.

Yes, thanks for the prompt responses to our queries.

> @sebb, I don't think Oliver is the only person who is facing this, I think I
> wrote somewhere here or on dev list that another person was facing this with
> 2.11 and tweeted about this.

PK, it would be interesting to know if they were also using S3.

> As discussed on dev list, we should:
> - write a wiki page about this setting
> 
> - Select a way to invalidate idle connections or connections that have run N
> requests (s3 case) for non 'clean' servers

I have just run a test against Amazon.com. They also have a server that fails
to send a Keep-Alive header. It seems to try to drop the connection after a
certain number of requests - by sending the following header:

Cneonction: close

Yes, it really is spelt that way!

What is odd is that it seems the connection still keeps going. In any case, I
suspect there is not a problem handling connection drop after N requests,
because that will be initiated by the server at the end of the response.

But I suppose it may still be worth considering recycling connections after a
certain number of requests.

The problem with the idle timeout is that it occurs at the start of a new
request, when disconnects are not currently expected. Maybe we can handle this
in JMeter, or it may need some help from HC.

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to