Hello,
thanks for your answer, but my dataset was an in-memory, that's why I
was confused with its behaviour. But I'll now recerate it and try to
repeat the issue to be absolutely sure.
Jaana
Rob Vesse kirjoitti 22.3.2021 13:36:
I assume that this is TDB 1?
It is possible you are encountering the scenario detailed at
https://jena.apache.org/documentation/tdb/faqs.html#fuseki-tdb-memory-leak
If the queries are sufficiently frequent the server may never be able
to flush the in-memory journal leading to continuous memory growth
over time. The above linked FAQ notes how you can diagnose if this is
indeed the case.
If you are impacted by this you would need to occasionally quiesce the
flow of queries to allow the server to flush the journal fully to disk
and free up the memory allocated to it.
Another alternative solution would be to switch over to using TDB 2
instead since that has a different on-disk memory structure that
avoids the need for updates to be kept in an in-memory journal. The
trade-off there is that it instead uses more disk space because there
are potentially multiple versions of disk on at any one time so you
would need to compact the database regularly -
https://jena.apache.org/documentation/tdb2/tdb2_admin.html - and/or
allocate more disk space to your cloud instances.
Hope this helps,
Rob
On 22/03/2021, 11:00, "jaa...@kolumbus.fi" <jaa...@kolumbus.fi> wrote:
Hello,
I'm running jena-fuseki 3.17.0 in azure cloud with memory-based
datasets. I haven't been updating any data into my datasets for one
week, but I just noticed that the memory consumption still keeps on
increasing (see attachment).
There's another service that queries data from that
jena-fuseki-instance
every minute via hypergraphql-interface. Could jena-fuseki somehow
cache
those requests causing memory consumption increase ?
If so, is there any means for preventing such caching ?
Br, Jaana
