On 8/31/2015 3:02 AM, Oleg Kalnichevski wrote:
> There likely to be something that sets SO_TIMEOUT to a lower value. You
> should be able to trace it by placing a breakpoint or good ol'
> System.out.println at this line:
> http://hc.apache.org/httpcomponents-core-4.4.x/httpcore/xref/org/apache/http/impl/BHttpConnectionBase.html#277
I added a line just before the existing line 277:
System.out.println("timeout debug: " + timeout);
I never did figure out how to pull httpcore into eclipse, so I initially
had a problem with tabs in the source from my editing escapade in
Notepad++. After I fixed that, I got a new jar, which I put into my
project and installed on the server. With lsof on the server, I can see
that the new jar (named httpcore-4.4.1-custom-with-println.jar) has been
used by Java, and the original file is NOT in use by that process.
I don't see anything in the file my start script creates by redirecting
stdout, though. If everything were working right, shouldn't I see
output from setting the timeout in my own code? I thought HttpClient
defaulted to zero (no timeout) if it wasn't specified ... was I wrong
about that? Is BHttpConnectionBase used by the deprecated
classes/methods as well as the suggested ones?
Right now it is looking like there's a problem with SolrJ setting the
socket timeout. I see code in our HttpClientUtil that does set the
timeout, so if this problem I'm suspecting is correct, either SolrJ is
doing it wrong or the deprecated classes aren't working right.
What thoughts do you have? Perhaps I should be creating the HttpClient
objects using HC methods directly instead of the SolrJ helper class.
Thanks,
Shawn
---------------------------------------------------------------------
To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org
For additional commands, e-mail: httpclient-users-h...@hc.apache.org
