> We set row cache too high, 10000 or so and now all our 6 nodes fail > with OOM. I believe that high row cache causes OOMs.
If you get into a fallen-and-cant-get-up scenario like this, I suggest a rolling restart of nodes, and on each node: (1) shut it down (2) removed saved caches (3) start it up (4) (quickly) use nodetool to decrease the cache sizes *on that particular node* Once that's done everywhere and everything is stable, a permanent and cluster-wide change of cache sizes can be done by a schema change as usual. -- / Peter Schuller (@scode, http://worldmodscode.wordpress.com)
