On Thu, Nov 13, 2014 at 03:19:17PM -0800, Mark Juric wrote: > Hello all, > I have an issue with the curl plugin. I'm collecting at 60 second intervals > from a web page on localhost, and it appears the connection is being held open > in between queries. For most situations this is okay, but I have a large > application installation that runs a "reaper" process to kill stuck Apache > threads. If I'm connected to a stuck thread that gets reaped, my connection > remains open and all future collections fail. The only solution is to > completely restart collectd. > > Is this a configurable option? The overhead of a new server connection every > 60 > seconds is negligible in my application, but the implications of losing data > collection even though the demon continues to run are troubling.
libcurl provides a way to accomplish this (the CURLOPT_FORBID_REUSE option) but it's not exposed by collectd. However, this shouldn't be necessary; libcurl will detect when a socket is closed and automatically open a new connection in that case. If the socket isn't actually closed, it's a different story. One of the timeout options or the TCP keepalive option should be enabled to catch that situation, but neither of those are exposed by collectd, either. >>> Dan _______________________________________________ collectd mailing list collectd@verplant.org http://mailman.verplant.org/listinfo/collectd