On 19/02/2013 14:37, Pavel Mateja wrote:
Can anyone give any clues as to what might be causing a timeout problem with mod_proxy being used as a reverse proxy to a local app server. This is using httpd 2.2.23 on ubuntu 10.04 LTS i686 I've (hopefully) eliminated the network itself as a source of the problem by testing this with the app server running on the same box as apache, using 127.0.0.1 as the network address in the proxy config. Even like this, I still get occasional '(70007)The timeout specified has expired: proxy: prefetch request body failed to 127.0.0.1:4000 (127.0.0.1)' messages in the logs. The app server shows the socket accept happening at the point the request is made, but then just sticks in a select() call until the timeout occurs (set to 5 seconds). No data is recieved from the apache mod_proxy. The apache logs then get the error message shown above added to them. The connection that mod_proxy is using is plain http, not https. This seems to happen maybe once or twice a day, no apparent pattern with system load, client type, url being requested etc. Is there anywhere that details the different error log messages that mod_proxy can generate, what they actually mean, and possible causes of them. I'm not even sure in this case if 'prefetch request body failed' relates to getting data from the app server or getting it from the web browser client (the error has 'to 127.0.0.1' in it which makes me think it relates to getting data from the app server) Is there any enhanced logging I can add to mod_proxy, for example to show it making the connection, packets sent, packets recieved etc. I just can't see why this timeout fails with mod_proxy not sending any data to the app server. Does mod proxy need the entire request from the browser before it forwards it on (what about requests with large post data? - though that is not the case in these errors). Can it have a long delay between making the connection to the app server and actually starting to transmit the request? Thanks for any hints/tips.Hi, do you use keepalive in your proxy? I had so set something like ProxyTimeout 41 Timeout 40 on proxy and Timeout 45 on backend because there is tiny race when backend closes connection just in the moment proxy tries to reuse that connection.
The proxy settings I'm using at the moment are: ProxyRequests Off ProxyErrorOverride On ProxyPreserveHost On ProxyTimeout 600 ProxyVia Off <Proxy balancer://appserver> BalancerMember http://127.0.0.1:4000 loadfactor=1 max=100 disablereuse=On ProxySet lbmethod=byrequests </Proxy>### Normally the balancer has more than one member, distributed over different servers but whilst trying to get to the bottom of this I've set it to just a single instance on the local box.
I've also tried disabling re-use by setting the nokeepalive and 1.0 request options in the <location> section that sets the reverse proxy:
<Location /tools/> SetEnv force-proxy-request-1.0 1 SetEnv proxy-nokeepalive 1 ProxyPass balancer://appserver/tools/ </Location>Using this the proxy does seem to be closing the connection immediately after each request. With the error I received today (the timeout shown above) it was at the start of a new connection, with no existing open connection to apache. The previous request from the same client had been almost 4 minutes earlier.
--------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx
|