On 10 December 2010 18:37, Peter Schuller <peter.schul...@infidyne.com> wrote: > Memory-mapped files will account for both virtual and, to the extent > that they are resident in memory, to the resident size of the process.
To clarify - in our storage-conf we have: <!-- This should consume a LOT less memory, and stop the JVM from ballooning immensely --> <DiskAccessMode>mmap_index_only</DiskAccessMode> I know it's a matter of degree, but 3GB in a month feels like a decent-sized ballooning. JNA is something I'd read briefly about a while back, but now it might be something I need to explore further. We're using Cassandra 0.6.6, and our Ubuntu version offers a packaged release of libjna 3.2.3-1 .. rumours on the Internets suggest a more recent version of jna may be required here. Though I'm also considering whether we just wear this as a known issue, do a rolling restart every 25 days (very low impact for us) and hold out until 0.7. We have a testing environment, but it's a pain to test solutions to problems that take a month to manifest. cheers, Jedd.
