We are currently having problems in out current production setup in solr. What we currently have is something like this:
- Solr 6.6.3 (cloud mode) - 10 threads for indexing - 900k total documents - 500 documents per batch So in each thread, the process will call a stored procedure with a lot of resultsets (1 main table and 8 sub tables) and after the db call, the application will assemble the documents based on the resultsets and then it will send it to solr for indexing. We are having errors such as heap space error in our indexing so we decided to lower the batch size to 50. The problem with this is that sometimes it really does not help since 1 document can contain 1000 child documents and it will still have the heap errors and indexing is generally slow everytime. So my question would be what approach should we have to resolve this kind of problem (will queue based indexing help? what are your indexing methods in your respective production environments?)? -- Sent from: http://lucene.472066.n3.nabble.com/Solr-User-f472068.html
