On Thu, Jan 26, 2012 at 3:55 AM, <d.guthmann_at_gmx.net> wrote:
> I think the reason is somewhere at the connection-specific configuration of apache in interaction with the behaviour of the subversion-client (which may has also connection-options...?).
>
> I suspect the KeepAlive-Handing... it seems that the subversion-client closes its connection between checking the different "externals" (which are on the same server) and the server keeps the connection opened and waits for further data. Until the connection is finally closed by timeout the server does NOT handle the request regarding the next "external".
>
> I don't have a clue why the server simply isn't accepting a second connection for the same client while the first connection is waiting for its timeout...
> I think one of the following options can affect this behaviour, but I'm still searching for an option which sounds like "ConnectionsPerClient":
> MaxRequestsPerChild
> KeepAliveTimeout
>
> Note: We are using a https-Connection.
I don't think apache has any connections-per-client restriction as
long as you have available children to accept them. Are you sure
this isn't some more subtle delay on new connections, like being
configured for reverse-DNS lookups for logging and timing out on the
DNS request? The proxy might be better at holding the connection with
keep-alives or maybe it is in an IP range where the reverse-lookup
either works or fails quickly with an icmp reject instead of a
timeout.
--
Les Mikesell
lesmikesell_at_gmail.com
Received on 2012-01-26 21:24:49 CET