28.09.2012 16:10, Suominen Osma wrote:
So I can cause Fuseki to run out of memory by just PUTting a little over
one million triples in total.
Again replying to myself...
I did some further testing. With Fuseki 0.2.3 from the Apache archive
[1] and using the same test setup, I can PUT the dataset at least 150
times with no problems. The memory usage (resident set size according to
top) of the Java process does grow slowly (1.4G at 50 requests, 2.1G at
100 requests, 2.7G at 150 requests) but at least Fuseki doesn't run out
of heap memory and stop working.
So it seems that the memory leak was introduced between the 0.2.3 and
0.2.4 releases of Fuseki.
The rising memory use is probably due to buffering/caching the growing
TDB database. It reached 1,9GB during my testing, which is a bit
strange, given that I'm just inserting the same 120k triples over and
over. But maybe that's just TDB transactions and journals in action.
-Osma
[1] http://archive.apache.org/dist/jena/binaries/
--
Osma Suominen | [email protected] | +358 40 5255 882
Aalto University, Department of Media Technology, Semantic Computing
Research Group
Room 2541, Otaniementie 17, Espoo, Finland; P.O. Box 15500, FI-00076
Aalto, Finland
