That looks like it. I've pushed the limits up to 65k and turned down the testing for now. Otherwise machines were dropping like flies. 

Thanks. 
Aaron

On 26 Aug, 2010,at 04:16 PM, Dan Washusen <dan@reactive.org> wrote:

Maybe you're seeing this: https://issues.apache.org/jira/browse/CASSANDRA-1416

On Thu, Aug 26, 2010 at 2:05 PM, Aaron Morton <aaron@thelastpickle.com> wrote:
Under 0.7.0 beta1 am seeing cassandra run out of files handles...

Caused by: java.io.FileNotFoundException: /local1/junkbox/cassandra/data/junkbox.wetafx.co.nz/ObjectIndex-e-31-Index.db (Too many open files)
        at java.ioRandomAccessFile.open(Native Method)
        at java.io.RandomAccessFile.<init>(RandomAccessFile.java:212)
        at java.io.RandomAccessFile.<init>(RandomAccessFile.java:98)
        at org.apache.cassandra.io.util.BufferedRandomAccessFile.<init>(BufferedRandomAccessFile.java:142)

If I look at the file descriptors for the process I can see it already has 1,958 for to the file 

sudo ls -l /proc/20862/fd | grep "ObjectIndex-e-31-Data.db" |  wc -l
1958

Out of a total of 2044. 

Other nodes in the cluster have a similar number of fd's - around 2k with the majority to one SSTable. 

I did not experience this under 0.6 so just checking if this sounds OK and I should just increase the number of handles or if it's a bug?

Thanks
Aaron