Renaming Solr mbean
Hi - I inquired about this some time ago, and learned a lot on the subject of JMX and Tomcat (and solr, too!), but never did find a definite answer. It doesn't look like anything JMX has been discussed on the list for some time, either. I've been toying around with JMX lately in Tomcat, and think its really neat how I can use JMX to query Solr instances with it. I can see the number of searches, min/max docs, warmup time, indice sizes, all kinds of neat stuff. In our setup, we run several instances of Solr under one instance of Tomcat. I simply rename the WAR to soemthing we use internally - solr-people, solr-connections, solr-companies, etc etc. This part works fine and lets us have, use, and maintain invidual instances. On the subject of JMX, however, I find that Solr always reports to the /solr mbean, regardless of the Solr instance name. Although this was not completely unexpected behavior, I had hoped that the mbean that solr would interact with was also named by the same name that the invidual instance of Solr was named. What I'm finding is every instance reporting to /solr, which skews my queries. Although I can somewhat predict which searcher is which, its not reliable enough to be able to associate statistics with our named versions of our indices. So in summary, I'm trying to figure out how to rename the mbean that Solr uses, specifically for JMX, to something a bit more predictable and isolated. Anyone know if/how I can do this? Thanks! -dant
Re: dose solr sopport distribute index storage ?
On 10/12/2009 10:49 AM, Chaitali Gupta wrote: Hi, How should we setup master and slaves in Solr? What configuration files and parameters should we need to change and how ? Thanks, Chaitali Hi - I think Shalin was pretty clear on that, it is documented very well at http://wiki.apache.org/solr/SolrReplication . I am responding, however, to explain something that took me a bit of time to wrap my brain around in the hopes that it helps you and perhaps some others. Solr in itself does not replicate. Instead, Solr relies on an underlying rsync setup to keep these indices sync'd throughout the collective. When you break it down, its simply rsync with a configuration file making all the nodes "aware" that they participate in this configuration. Wrap a cron around this between all the nodes, and they simply replicate raw data from one "master" to one or more slave. I would suggest reading up on how snapshots are preformed and how the log files are created/what they do. Of course it would benefit you to know the ins and outs of all the elements that help Solr replicate, but its been my experience that most of it has to do with those particular items. Thanks -dant
Extended stats via JMX
Hi - This is my first night working with JMX, particularly for the purpose of querying Solr statistics running under Tomcat. Before, I was trying to xpath to stats.jsp which just felt dirty. I gotta say - I think this is pretty neat. Right now, being inexperienced with JMX and all, I was wondering if there was a way to pull all Solr-specific items out of there. I see some general counters pertaining to each of my Solr instances, but nothing along the lines of lookups, hits, hit ratios, and the like. They're all more Tomcat-centric - memory usage etc etc. That's fine, too, but the whole point of this exercise is to get instance-specific statistics regarding Solr. Is this information exposed via JMX under Solr? Can I pull a list somewhere of all items that find their way through JMX to be seen from an external source? Thanks in advance -dant
RE: Extended stats via JMX
-Original Message- From: Matthew Runo [mailto:mr...@zappos.com] Sent: Thursday, February 25, 2010 12:18 PM To: solr-user@lucene.apache.org Subject: Re: Extended stats via JMX https://issues.apache.org/jira/browse/SOLR-1750 might help you, since I don't think that all of stats.jsp is exposed via MBeans. I could be wrong about that though.. (apologies, our solr servers are firewalled and I can't connect via JMX at the moment) Thanks for your time! Matthew Runo Software Engineer, Zappos.com mr...@zappos.com - 702-943-7833 Hi, Matthew - Looks like Shalin confirmed that those values can in fact be found inside the RequestHandler's MBean (thanks, Shalin). Thanks for getting me going in the right direction. I appreciate it. Thanks -dant
Monitoring via JMX; changing mbean names?
Hi - I've not been able to find a way to change the names of mbeans exposed from solr via JMX. Is this even possible? Am I just plain doing it wrong? For example, when running multiple instances of solr in the same Tomcat instance, each has an associated searc...@1234567 mbean. Alright, I expect that. However, some values that I'm looking for (such as avgRequestsPerSecond, avgTimePerRequest) are all located under the solr/standard mbean. I'm not sure which actual instance of solr this value actually represents. Now, to be fair, I'm very new to using JMX for anything. My conclusions have involved a lot of speculation, and honestly, some good guesses. Maybe it would be wise to have a separate JMX server for each instance of Solr? I don't know. The overall goal is to be able to monitor response times and memory usage from Solr, and even further each instance of Solr. I think JMX is the solution, unless someone else would be so kind as to provide some other recommendations. Thanks! -dant
