> This is a crash with Cassandra 0.7.0-beta3 but I am pretty sure I also have > got this on a rc1 server too ( We are dropping and recreating keyspaces when > this happens) > > > > I do have the JNA enabled but it looks like it is still calling an external > program – presumably this would cause it to ask for the same memory as the > parent process and hence die?
Yes. Either enable overcommit in the Linux kernel, or use a sufficiently new cassandra with working JNA to avoid use of fork+exec for running ln. (I forget when the change wen in so not sure whether beta3/rc1 is recent enough for the ln fix.) -- / Peter Schuller