We will now use the cassandra-stress tool to write several thousand records to our node. Execute the following command in your original terminal:

cassandra-stress write no-warmup n=250000 -port native=9042 -rate threads=1

Open the second terminal so you can watch as cassandra-stress writes 250,000 rows to your node.

There are a few things to watch out for while cassandra-stress inserts keys:

• The total size will continue to increase.
• The timestamp will change for the current segment being written.
• You may get additional commit log files as well.

When cassandra-stress completes, terminate the watch by pressing CTRL-C.

echo "Sending a Ctrl-C"

Execute the following nodetool command:

nodetool cfstats keyspace1.standard1

cassandra-stress created the keyspace1.standard1 table and populated its data. cfstats gives you column family stats (column family is a deprecated term for a table).

Notice the "Write Count" matches the number of rows we told cassandra-stress to insert. cfstats also reports the number of SSTables, space used, and bloom filter statistics.

Also note the Memtable statistics.

Execute the following nodetool command which will flush the memtable contents to disk.

nodetool flush

Now check the table stats again by executing this command:

nodetool cfstats keyspace1.standard1

Note that the memtable statistics are zeroed out because we flushed the previous memtable to disk.

Shut down your node, delete the logs/system.log file, restart your node, then search for CommitLog.java in the new logs/system.log file. You may see lines reporting replays.

TODO add commands for this

If there were no commit log segments found during startup, no replay needs to be done. If Cassandra finds commit log files, it will replay the mutations in those files into memtables and then flush the memtables to disk.

Write Path Quiz