Check that read_repair_chance on the CF's is 0.1, not the old 1.0 Wait at least 10 minutes for the DynamicSnitch to re-calculate.
Use the org.apache.cassandra.db:type=DynamicEndpointSnitch MBean to see what scores it has given the nodes. Cheers ----------------- Aaron Morton Freelance Cassandra Consultant New Zealand @aaronmorton http://www.thelastpickle.com On 8/03/2013, at 11:40 AM, Edward Capriolo <edlinuxg...@gmail.com> wrote: > dynamic_snitch=true is the default. So it is usually on "wrapping" other > snitches. I have found several scenarios where it does not work exactly as > your would expect. > > On Fri, Mar 8, 2013 at 2:26 PM, Hiller, Dean <dean.hil...@nrel.gov> wrote: > Our test setup > > 4 nodes, RF=3, reads at CL=QUOROM and we tried CL=TWO > Tell the network card to slow down every packet on node 2 > After fixing astyanax to not go to node 2 anymore, we are still seeing > cassandra have issues as it seems to be involving node 2 somehow. If we take > node 2 down, it all speeds back up. > > We are trying to get this working such that a slow node in cassandra does not > impact our customers. > > We are in 1.2.2 and added the following properties….(our properties show > PropertyFileSnitch though I see the keyspace has > org.apache.cassandra.locator.SimpleStrategy set probably because it was > created through a tool instead of CLI…shucks)….anyways, I still expected > dynamic snitch to work…. > > # controls how often to perform the more expensive part of host score > # calculation > dynamic_snitch: true > dynamic_snitch_update_interval_in_ms: 100 > # controls how often to reset all host scores, allowing a bad host to > # possibly recover > dynamic_snitch_reset_interval_in_ms: 600000 > # if set greater than zero and read_repair_chance is < 1.0, this will allow > # 'pinning' of replicas to hosts in order to increase cache capacity. > # The badness threshold will control how much worse the pinned host has to be > # before the dynamic snitch will prefer other replicas over it. This is > # expressed as a double which represents a percentage. Thus, a value of > # 0.2 means Cassandra would continue to prefer the static snitch values > # until the pinned host was 20% worse than the fastest. > dynamic_snitch_badness_threshold: 0.1 > > Any help appreciated, > Thanks, > Dean >
