Performance problems during import / conversion

[Flavio 123]

Hello

We import via Java or apoc.load.jdbc about 100'000 nodes to Neo4J 3.4.0. The import script consists of several steps. Among other things, the original graph is converted into an extended graph, which consists of about 500'000 nodes at the end. 

The complete import takes about 3 minutes. But, we had to implement sleeps of 20 seconds between the steps. Otherwise the import takes a multiple of 3 minutes. 

* Indexes are created (with "CREATE INDEX ON :NodeXY(fieldXyId)") .

* Indexes are ONLINE (tested with "CALL db.awaitIndex(':NodeXY(fieldXyId)')"

What could be the reason for the needing of the sleeps? Are the indexes not online? How can we check if Neo is ready for next queries/conversions?

Thanks for your help!

[Flavio 123]

I have posted the question also on Stackoverflow: https://stackoverflow.com/questions/51038362/performance-problems-during-neo4j-import-conversion-with-multiple-steps

Still is any feedback welcom, since the QM will not accept sleeps in our code...

[George & Sherin]

Are you really using those indexes in the load process. If not I would suggest you to do the data load and do the index creation at the end. 

George S

On Tue, Jun 26, 2018 at 2:04 PM, Flavio 123 <fsad...@gmail.com> wrote:

I have posted the question also on Stackoverflow: https://stackoverflow.com/questions/51038362/performance-problems-during-neo4j-import-conversion-with-multiple-steps

Still is any feedback welcom, since the QM will not accept sleeps in our code...

--
You received this message because you are subscribed to the Google Groups "Neo4j" group.
To unsubscribe from this group and stop receiving emails from it, send an email to neo4j+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.