Salut,
Kristian Rink wrote:
> Salut Jeanfrancois, all;
>
> ... here we go again, some more reports on that:
>
> Kristian Rink schrieb:
> [...]
>> that. So far I am into building a testbed installment running apache 2.2
>> (our productive system is 2.0) which I would prefer in the future
>> because (asides from the usual updates) it provides the mod_proxy
>> balancing features which I want to make our glassfish clustered;
> [...]
>
> Been running apache 2.2 / mod_proxy now for two days in a productive
> environment. It seems more stable than the previous (2.0) setup, and
> considering the fact it have been two "ordinary" work days in terms of
> concurrent user count, it seems effectively the amount of
> ClientAbortExceptions to hit my error log has somewhat reduced (somewhere
> next to 100 in these two days, whereas just a handful of people actually
> complained about being unable to recieve / fully download files). Things I
> have noticed about this:
>
> - In some cases, it seems some given resources are more troublesome than
> others. One of the URLs accessed by external users is supposed to send back
> an 8..10 MB .xls file which is causing serious trouble to quite an amount of
> people.
This is interesting.
In these cases, it also seems a "final" thing to these users -
> either they manage to completely download the resource in question
> immediately, or they will always fail. In some cases, I definitely blame
> this to user side network connectivity issues but at least on our system I
> see the exception and I am stunned to see their browsers reporting "the
> website is not available or not found"...
It might also be caused by mod_proxy not able to buffer the file
completely in memory? Is mod_proxy configured to buffer files from
GlassFish or flush it on the fly? I think you just found something
interesting. Can you set the
<keep-alive .. timeout-in-seconds="0"/>
in domain.xml. What it will do is to force mod_proxy to not re-use the
connection, but instead open a new connection on every request. This is
not the perfect configuration (increase network latency), but at least
mod_proxy will not close a connection when GlasFish is writing some bytes.
>
> - For what I see on the apache2 proxy logs, whenever another one of our
> backend systems (a proprietary DMS web access...) causes problems, the
> "connection reset by peer" is also logged in apache2's error.log file so I
> assume that, given in the ClientAbortException the apache2 might be cause of
> the problem, it might also be reported there?
Reporting to the httpd mailing list would be the best place...but keep
us updated about our finding....that help a lot the GlassFish community
which uses mod_proxy!
>
> Still clutching at straws, I think... :/
> Cheers & best regards, have a good week everyone...
A+
-- Jeanfrancois
> Kristian
>
>