>> The stack trace doesn't make sense relative to what I get checking out >> 0.6.6. Are you *sure* this is 0.6.6, without patches or other changes? > > Oh, sorry, the original poster of this thread was/is actually using 0.6, I am > (as mentioned in other posts) actually on 0.7rc2. Sorry that I didn't point > this out once again.
Sorry, my bad. I tend to catch up on the ML:s during transportation to/from work and is in 'fast sifting' mode ;) Sometimes multiple people from the same group/org/company posts int he same thread about the same issue and I thought this was the case here. In that case, based on the strack trace, I wonder if you're hitting what I was hitting just yesterday/earlier today: https://issues.apache.org/jira/browse/CASSANDRA-1860 Which is suspected (currently being tested that it's gone with updated 0.7) to be due to: https://issues.apache.org/jira/browse/CASSANDRA-1847 If the corruption happens to be slightly different in your case, maybe it just ends up trying to allocate a large array instead of the failure mode I got (EOF or bad cfId) on the receiving end. Can you try to reproduce with the latest 0.7 branch as of today? It seems very similar in nature at least based on the latest stack trace. -- / Peter Schuller
