If you're a unix shell scripting wiz, here are a few strategies. Tail the logfile and filter for the string 'QTime'. The number is the very last string in the line. So, strip the text between the timestamp and the number- sort by the timestamp first and the number second. Now grab the first qtime for each timestamp. I don't know a command for this. This gives you the longest query time for each second.
As a separate trick: tail the logfile and filter for QTime. Then, strip out all text after the time. Now you have a stream of lines with a timestamp. Run this through 'uniq -c' and voila! you get the queries per second for each timestamp. On Sat, Feb 12, 2011 at 1:51 AM, Gora Mohanty <g...@mimirtech.com> wrote: > On Sat, Feb 12, 2011 at 4:54 AM, Stijn Vanhoorelbeke > <stijn.vanhoorelb...@gmail.com> wrote: > [...] >> Can you access this URL from a web browser (tried but doesn't work ) ? Or >> must this used in jConsole / custom made java program. > > Please try http://localhost:8983/solr/admin/stats.jsp (change hostname/port as > needed). > >> Could you please point me to a good guide to implement this JMX stuff, cause >> I'm a newbie for JMX. > > The easiest way to get access to JMX is indeed a Java console, like jconsole. > There are various open-source JMX clients available, but we could find none > that met our needs, and were being actively maintained. We have been > toying with the idea of a JMX client that offers a REST API to Solr MBeans > (or even to any generic MBeans). This would be a more natural interface for > people used to web development. > > Regards, > Gora > -- Lance Norskog goks...@gmail.com
