We've been running 7.2.1 at work for a while now and it's been running very well. We saw some performance improvements, but I would attribute most of that to the newer instance types we used in the upgrade. Didn't see any major performance regressions for our workload.
A couple of things to think about: * You might want to wait for 7.3 since it is at RC now and there is an annoying bug that is fixed when reloading configuration doesn't propagate to PULL replicas. May or may not matter to you if you stay with NRT replicas. * If you are concerned with performance, try running some prod queries against a test cluster. I think you'll be fine and probably should think about any relevance changes that 7 introduces. There are things like "split on whitespace" etc that are behavioural changes that may or may not matter for you. For what it's worth, we kept the defaults, upgraded our analysis chains to undeprecated versions and didn't have any problems. On Tue, 27 Mar 2018 at 03:17, Walter Underwood <wun...@wunderwood.org> wrote: > If you are running 6.4.1, you will see a big speedup when going to a later > version. The metrics code caused a serious performance problem. > > https://issues.apache.org/jira/browse/SOLR-10130 > > wunder > Walter Underwood > wun...@wunderwood.org > http://observer.wunderwood.org/ (my blog) > > > On Mar 25, 2018, at 8:55 PM, S G <sg.online.em...@gmail.com> wrote: > > > > Hi, > > > > Solr 7 has been out for about 6 months now. (Sep-2017 to Mar-2018) > > We are planning some major upgrades from 6.2 and 6.4 versions of Solr > and I > > wanted to see how is Solr 7 looking in terms of stability and > performance. > > (Have seen http://lucene.apache.org/solr/news.html but some real users' > > experience would be nice) > > > > 1) Has anyone encountered major stability issues that made them move back > > to 6.x version? > > > > 2) Did anyone see more than 10% change in performance (good or bad)? I > know > > about https://issues.apache.org/jira/browse/SOLR-11078 and wish trie > fields > > were still kept in schema until point fields completely got over the > > performance issue. > > > > Thanks > > SG > >
