90+% of the time when I see this question it’s a problem with the client not being able to push docs at Solr fast enough. This is particularly true if databases are involved.
In addition to Jörn’s comment, I’d ask whether your Solr CPUs are running flat out. If your CPUs aren’t maxed out, you need to send more docs to Solr. It looks like you have a single indexing Solrj client. The bottleneck might be there, i.e. at 30 threads the _client_ is acquiring docs as fast as it can, due to CPU utilization, disk I/O, whatever and Solr is just idling along. Certainly Jörn’s comment about batching docs is germane. Here’s some quantification: https://lucidworks.com/2015/10/05/really-batch-updates-solr-2/ Best, Erick > On Jun 2, 2019, at 1:15 AM, Jörn Franke <jornfra...@gmail.com> wrote: > > And to send multiple documents in one add document step > >> Am 02.06.2019 um 08:45 schrieb calamita.agost...@libero.it.invalid: >> >> >> >> Ho all, >> I've an ingestion application that reads many files and indexes records >> in a SolrCloud 7.4 with SolrJ using addDocument. >> The application is multithreading, every thread reads a file and send >> addDocument. >> I see that from 10 to 30 threads executing addDocument in parallel, number >> of files indexed increase from 120 to 360 files per minute. >> If I use 50 threads, the number of files per minute remain the same. >> Is there a parameter to customize in Solr to increase parallelism relative >> to ingestion with addDocument http interface ? >> Thanks. >> -- >> Inviato da Libero Mail per Android
