Hi all .. I've been meaning to ask this question and now it's kind of urgent ..
We are running Dataverse 6.11 on Payara 7 with the standard Apache + mod_proxy_ajp + Shibboleth configuration from the installation guide. We are occasionally seeing mod_proxy_ajp BufferOverflowException / ajp_marshal_into_msgb: Error appending the header value errors during Shibboleth-authenticated requests, resulting in a 400 Bad Request before the request reaches Payara. Our institution's IdP is known to send relatively large SAML-derived headers, and we've previously had to work around header-size issues by stripping cookies on /shib.xhtml.
Has anyone successfully migrated a Dataverse 6.x installation from mod_proxy_ajp to mod_proxy_http with Shibboleth? If so, were there any Dataverse- or Shibboleth-specific configuration changes required beyond updating the Apache proxy configuration?
In the guides, we definitely promote using AJP, like this:
ProxyPass / ajp://localhost:8009/
But long ago I think I played around with...
ProxyPass / http://localhost:8080/
... and I seem to remember it working. Maybe give it a try on a test server?
It looks like I took some notes over a decade ago at #2294 if that helps! :sweat_smile:
Some possible security implications? I'm not sure.
Ah yea, if I'm reading that thread right, it seems that method was not really suggested .. is not supported and seems insecure .. so I likely won't go that way .. need to figure out some other way to resolve this header issue .. i did find there is a ProxyIOBufferSize directive that I can add to the apache config .. will try that. Thanks!
This is the first I've heard of large SAML-derived headers and having to modify shib.xhtml because of them.
I wonder if you can handle this somehow in your Shibboleth config.
For example, @Kirรกly Pรฉter was able to remap an attribute from one to another (e.g. from uid to eppn) by specifying a rule in attribute-map.xml.
Of course, this may be completely different from your situation.
If you haven't already, you might want to ask on the shib-users mailing list. I haven't posted in a while but they've always been friendly and responsive.
I didn't modify shib.html .. I just added RequestHeader unset Cookie to Location /shib.html in the apache config. I thought that would help but apparently it didn't. Yea I'll ask on the shib-users list .. I've asked things there before. I'm not sure they are SAML headers this is all just a guess, that's why I posted here, just wondering if anyone else has run into this.
Sure. Maybe @Don Sizemore knows.
I did find that Apache was still using the default AJP buffer. After setting ProxyIOBufferSize 65536, the mod_proxy_ajp debug logs now show data to read (max 65530 at 4) instead of the previous 8 KB ?? limit. Leaving this in place to see if it resolves the intermittent BufferOverflowException during shib account creation. :fingers_crossed:
That would be sweet!
I've never encountered this, no. What on earth does your IdP send that's larger than 8kb?
the more i look into it the more i think it is not just shib headers, but cookies, shib information, and other request metadata so that the overall authenticated request from ASU headers is too large .. but as I said I'm just guessing at this .. idk
we have a similar issue when users make a lot of requests to our vendor e-resources .. after a lot of searching they start to get a "request header too large" error from those vendors because of the org headers added
Last updated: Nov 02 2026 at 07:01 UTC