Hi,

2GB for initial / max memory.
Joseki is the only application deployed.
I only ran 35 queries (consecutively) before memory ran out.
These queries were quite complex and included ARQ non-SPARQL capabilities. They were pretty much exactly the same, except for a FILTER part.

Anyway, in my case, query performance does not seem to be crippled by the disabling of statement caching, so I will probably leave it disabled.

Regards,
Julien

Hi,

I had the issue even with only one connection in the connection pool (defined in joseki-config.ttl).

Regarding the queries, I am sorry, but I can't discuss them much due to the confidentiality of the matter.

However I can say that the SPARQL query that was mainly used involved :
- ORACLE_SEM_FS_NS settings : DOP=8, timeout and qid
- (sometimes) a nested SELECT
- an UNION between several graph patterns, each of them using FILTERs
- apf:splitURI and fn:substring
- around 35 triple patterns
- several FILTERs (REGEX, ISURI, etc)
- finally, ORDERing.

Each query resulted in thousands of lines.

Model is a ~17,000,000-triple raw (no inference, no virtual model, etc) RDF model.
Joseki settings were pretty much the default one. As for the datasource.

To be honest, I doubt that statement caching would be of any use in my case since it is very unlikely that the same queries will be sent several times.
As I said, disabling it did not affect the query performance much.

It is clearly not really an issue for me (I mainly talked about it here in case someone else had OoMs with Joseki and was trying to find out why), so if I am the only isolated user who encountered this, it might not be useful to investigate further...

Thanks,

Regards,
Julien