You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@jena.apache.org by "Neubert, Joachim" <J....@zbw.eu> on 2014/12/01 17:42:04 UTC

RE: Query aborts with 502 behind reverse proxy

Hi Andy,

I'm now quite sure that the timeout problem is not related to our server itself. But there is indeed a load balancer involved (and perhaps other components out of my scope), and if I circumvent that, the query works even with the reverse proxy. 

Not sure if _that_ issue can be fixed. But thank you very much for the hint.

Cheers, Joachim


> -----Original Message-----
> From: Andy Seaborne [mailto:andy@apache.org]
> Sent: Wednesday, November 26, 2014 11:11 AM
> To: users@jena.apache.org
> Subject: Re: Query aborts with 502 behind reverse proxy
> 
> On 26/11/14 07:12, Neubert, Joachim wrote:
> > I use Fuseki behind a apache reverse proxy. A query which takes about 90
> seconds when put directly to Jetty, aborts invariably after 60 seconds with
> an "The proxy server received an invalid response from an upstream server.
> The proxy server could not handle the request" message by apache.
> >
> > In the apache config, I set all values known to me to 10 minutes:
> >
> > Keepalive On
> > Timeout 600
> > ProxyTimeout 600
> >
> > ProxyPass         /beta/sparql/thesozv/query
> http://127.0.0.1:3030/thesozv/query timeout=600 Keepalive=On
> > ProxyPassReverse  /beta/sparql/thesozv/query
> > http://127.0.0.1:3030/thesozv/query
> >
> > Jetty configuration is a bit a black box for me. However I found that there
> seems to be a 60 second default timeout in the ProxyServlet (init parameter
> "timeout" - http://www.eclipse.org/jetty/documentation/current/proxy-
> servlet.html).
> >
> > I'm not sure if this setting is involved at all, and how I could set it, or what
> could be wrong otherwise. Help would be greatly appreciated.
> >
> > Cheers, Joachim
> >
> 
> Joachim,
> 
> Nothing jumps out at me.  We run with a setup that allows queries longer
> than 60s.  We rely on the Fuseki timeout and bump other timeouts above
> that so only one system is causing a timeout.
> 
> The httpd default timeout is 300s IIRC.
> 
> We don't set Keepalive=On. We don't have a ProxyPassReverse directive.
> 
> We do set max= to stop too many queries at one time.
> 
> Fuseki does not use ProxyServlet.
> 
> Aside
> 
> An AWS-ism: the load balancers default to a 60s timeout.  Nowadays (since
> July this year, 2014) you can increase that to 5 mins in the UI.
> 
> 	Andy