Hi, I don't have a curl bug, but I have a question about usage that I thought someone here might be able to help me with.

I've built libcurl into a multi-threaded server that has to do a LOT of web spidering. Things were
going great until we started stress-testing.


When I run for a long time getting urls constantly (like 40 hrs), I start to get more and more curl handles coming back null when I call curl_easy_init().

I'm running this test on red hat 7.2.

I think I'm probably running out of either sockets or file descriptors. I don't have a ton of linux experience (this problem doesn't happen to me on windows), and I was wondering if someone
here could suggest approaches I should take to this problem.

thanks,
Michael