Topic: Apache Proxy errors AH01102 + AH00898 - Request too long

[Renaud SALVIGNAC]

Hello,

I have an error on Apache + proxy with balancer.

When an url exceeds a certain length, I have an Proxy error 502 :

[tangent]

You don't say what your upstream load balanced servers are running, but presume you've checked to see what corresponding error is being recorded there?

Your request URL alone is over 1K in length, and request headers in general have grown in size over time (multiple cookies, authentication tokens, X-headers, etc.), so is it possible you've exceeded some buffer size for some of your client requests?

In the past I've seen this problem with Tomcat backend servers, where we needed to increase maxHttpHeaderSize in server.xml. Other backend services will equally have some finite defaults.

Also, at the Apache level, check out ProxyIOBufferSize which defaults to 8192 bytes.

[Renaud SALVIGNAC]

Hello,

Sorry. I have an other apache serveur behind.

If I test directly the url with local IP in replacement of domain, it works.

Errors are logged on Proxy server, not on server behind.

I've already tested ProxyIOBufferSize directive.
It doesn't change anything.

[James Blond]

You may set

[Renaud SALVIGNAC]

Also tested without any result...

[James Blond]

Please post your reverse proxy config.

[Renaud SALVIGNAC]

Below the config

[tangent]

Looking at your proxy configuration, and the error log detail in the original post, I notice the error relates to remote server 192.168.202.41:80 whereas your balancer member IP's are 192.168.202.101, 102 and 103. Is this relevant, or did you change the balancer IP's for the posted config?

Otherwise, if you've already tried increasing the buffer sizes, I suspect the problem lies somewhere with the proxied headers being fragmented / truncated, or relates to compression, but to analyse this further believe you need more detail to crack where the exact problem lies.

There are two things I'd try next here.

Firstly try removing the Accept-Encoding request header, so that the back end servers send an uncompressed response to Apache. Put the following in your virtual host block or global configuration (if acceptable).

[Renaud SALVIGNAC]

We have 1 proxy and 3 environments (prod / dev /beta).
The configuration given is about Prod.

IP address 192.168.202.41 is for dev environment.

I've added "RequestHeader ..." but it doesn't change anything.

I'm checking with tcpdump but not simple to analyze.

[tangent]

Ok, so your one proxy covers multiple environments. Presumably, this error occurs across all three, albeit inconsistently?

Assuming you can narrow the problem to say the dev environment, you should be able to craft a filter for tcpdump to restrict the captured traffic (written to a pcap file). I'd personally just capture and analyse the traffic between the proxy and corresponding/potential back-end servers (hence the recommendation to remove the Accept-Encoding request header, so the proxied content isn't compressed).

I tend to transfer the pcap files over to a Windows platform, where it's easier to analyse them with Wireshark.

Let us know how you get on.

[Renaud SALVIGNAC]

Yes it's the same on each environment.

I try your method and give you feedback.

[Renaud SALVIGNAC]

We found a workaround.

We use proxy server in "normal web mode" for one subdomain.
We store parameters then we redirect to other script that recover the paramaters.

Thanks for your help

[tangent]

Not sure I fully follow your explantion, but am glad you've resolved the problem with a workaround.

Apache Lounge has numerous subscribers with many years of experience. Don't hesitate to post queries for help, or equally your expertise in solving problems.

[Renaud SALVIGNAC]

I will not miss it.

Thanks for your advices.