Ah ok. Sorry for the RTFM fail John ;-). For my test with a single node, batch would make sense if I needed better durability but with a cluster it's less of a concern with replication.
Thanks. -Mark On Sat, Apr 10, 2010 at 11:49 AM, Jonathan Ellis <jbel...@gmail.com> wrote: > http://wiki.apache.org/cassandra/Durability > > On Sat, Apr 10, 2010 at 10:38 AM, Mark Greene <green...@gmail.com> wrote: > > Hi, > > I'm testing out failover for 0.6.0-RC1 and seeing varied behavior in > > Cassandra's ability to replay the commit log after a forced failure. > > My test is this: > > 1) Run ./cassandra -f > > 2) Insert a value through the CLI and immediately force a shutdown of > > cassandra after I see the "Value inserted" confirmation from the CLI. > > If I shutdown cassandra quickly, within a second or two of seeing "Value > > Inserted" the data seems to be lost after starting cassandra back up. I > did > > a nodetool repair and flush as well just in case that may have had any > > affect. > > Conversely, if I waited about 5 seconds before shutting down cassandra, > then > > started it again, I was able to retrieve the data from the CLI. > > > > Any ideas? > > Thanks in advance, > > Mark >
