Hi Shawn, Thank you for your feedback, I much appreciate your inputs. I don't have strong requirements regarding structuring the data: do you think I could use a single, relatively large collection with some discriminator field instead of multiple thousands of separate collections?
Thanks, Peter 2016-05-12 20:30 GMT+02:00 Shawn Heisey <apa...@elyograg.org>: > On 5/12/2016 9:08 AM, Horváth Péter Gergely wrote: > > As part of benchmark, I attempted to create about 2500 collections to > > see how well that would work for us. Unfortunately, the experiment > > yielded some disappointing results, after about 2000 being created > > SolR got hung; REST requests started failing. I found the following in > > the logs: > > Solr will not handle that many collections very well. You're pushing > the boundaries of scalability. See this issue that I created: > > https://issues.apache.org/jira/browse/SOLR-7191 > > Are you creating the collections sequentially, or running multiple > CREATE actions simultaneously? Sequentially, where you wait for a > previous CREATE to complete before executing another one, is strongly > advised. > > SolrCloud starts to have serious problems when you create a lot of > collections. We are aware of the scalability issues, but they are not > easy to fix. > > Thanks, > Shawn > >
