httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From eric <spamsin...@gmail.com>
Subject Re: [users@httpd] Server Drops Client Intermittently
Date Tue, 19 Aug 2008 16:18:24 GMT
On Tue, Aug 19, 2008 at 5:21 PM, Nick Kew <nick@webthing.com> wrote:
> On Tue, 19 Aug 2008 17:02:08 +0200
> eric <spamsink42@gmail.com> wrote:
>>
>> [Tue Aug 19 16:58:50 2008] [debug] mod_deflate.c(749): [client
>> 99.999.999.99] Zlib: Inflated 407475 to 12475739 : URL /ws/myapp
>> [Tue Aug 19 16:58:50 2008] [warn] (20014)Internal error: mod_fcgid:
>> can't get data from http client
>
> Is 12475739 the correct uncompressed size of the request?

Yes.  Earlier I mentioned that the threshold for problems is a message
size of ~10MB compressed, I was mistaken, it's ~10MB uncompressed.

> If so, looks like it uncompresses OK, so what happens next?

I'm not sure.  I have started apache with -X and attached gdb to it
but I'm unable to determine much.  The source code is not present in
the production environment, I have tried without success to acquire
the source code which corresponds to the installed binaries for
httpd2, mod_fcgid, etc, is there a standard procedure with SUSE to do
this?

> Could it be violating some hard limit in fcgid or in your ipc,
> or some configured limit?

I have investigated such questions for several days, I haven't been
able to establish anything definitive.  We have that one site where
everything always works.  Here it succeeds intermittently.  Even when
the query size is increased to 100MB (before compression), it still
succeeds at least sometimes from every site.  Plus the failure returns
in 5 seconds which is well within all of the configurable timeouts
that I can see.  These facts seem to rule out many of the likely
explanations for limits and timeouts at whatever point in the chain.

>> I forgot to mention, a failed iteration returns in about five seconds,
>> a successful one in ten.
>
> So the overhead of using AN Other CGI implementation is going
> to be negligible, if necessary.

When I get desperate enough I will reconfigure the application as a
standard CGI application, at least to see if that fixes the problem.

> What are you doing with mod_headers there?

Nothing explicit - whatever is happening with mod_headers is just a
result of a mostly default implementation of Apache, FastCGI etc.

> It shouldn't make
> any difference, but if "RequestHeader unset Content-Length"
> or any other Content-* header makes a difference, you might've
> got a bug.

I just quickly tried it with "RequestHeader unset Content-Length"
which caused the application to fail altogether - what does that tell
us?

Regards,
Eric

---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
   "   from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org


Mime
View raw message