>> For debugging purposes you may want to switch Cassandra to "standard" >> IO mode instead of mmap. This will have a performance-penalty, but the >> virtual/resident sizes won't be polluted with mmap():ed data. > > Already did so. It *seems* to run more stable, but it's still far off from > being stable.
Sorry, that was written under confusion. I meant for the original poster and his monitoring of virtual/resident memory and the fact that he seemed to have a legitimate OS level out of memory condition. In other words, I recommend that to the OP just to remove one variable in terms of monitoring what is consuming memory. -- / Peter Schuller