I'm having a problem with HTTP uploads. When I try large file uploads (>100MB ) via HTTP they start and then eventually fail. I am using a two server SuSE linux cluster running Tomcat 5. I've temp. disabled one of the servers in the cluster to simplify troubleshooting. If I go to the server directly, the upload always works.
I'm thinking maybe this has something to do with HTTP keepalives? Can someone recommend a troubleshooting methodology to narrow it down?
Are you using POST to upload the files? There is a POST timeout of 60 seconds on the DX to stop potential resource starvation as the DX buffers the client data, do you find that the upload stops after 60 seconds? This timeout can be controlled globally or on a per cluster basis, to view the settings use:
% show server factory h pt Factory h pt: 60
% show cluster <cluster name> factory h pt factory h pt: global
you can change these with 'set' commands, followed by a 'write' e.g.
% set server factory h pt 200
(*) % write
If this is not the problem, the first thing to try is to test with a SLB group, to confirm the problem only happens with a cluster. If that works you could use apprules to drop to forwarder mode for the URL accessed to upload the content. This makes the connection not use the cluster code within the DX, so it then is like connecting through an SLB group e.g.
RTH: url equals "/upload.cgi" then forward
If this does not work then I would suspect the problem is either buffers getting full or a connection timeout. Does the DX report anything in the System log ('show log system')?