As you said, the auto_bootstrap setting was turned on. 

Well I was talking about the 'auto_snapshot' ;-). I understand that's what you meant to say.

This command seems to apply only to one node. Can it be applied cluster-wide? Or should I run this command on each node?

Indeed, 'nodetool clearsnapshot' is only for the node where you run the command, like most of the nodetool commands (repair is a bit specific).

C*heers,
-----------------------
Alain Rodriguez - @arodream - alain@thelastpickle.com
France / Spain

The Last Pickle - Apache Cassandra Consulting
http://www.thelastpickle.com

2018-01-15 1:56 GMT+00:00 Eunsu Kim <eunsu.bill23@gmail.com>:
Thank you for your response.  As you said, the auto_bootstrap setting was turned on. 
The actual data was deleted with the 'nodetool clearsnapshot' command.
This command seems to apply only to one node. Can it be applied cluster-wide? Or should I run this command on each node?



On 12 Jan 2018, at 8:10 PM, Alain RODRIGUEZ <arodrime@gmail.com> wrote:

Hello,

However, the actual size of the data directory did not decrease at all. Disk Load monitored by JMX has been decreased.

This sounds like 'auto_snapshot' is enabled. This option will trigger a snapshot before any table drop / truncate to prevent user mistakes mostly. Then the data is removed but as it is still referenced by the snapshot (hard link), space cannot be freed.

Running 'nodetool clearsnapshot' should help reducing the dataset size in this situation.


The client fails to establish a connection and I see the following exceptions in the Cassandra logs. 
org.apache.cassandra.db.UnknownColumnFamilyException: Couldn't find table for cfId…

This does not look like a failed connection to me but rather a try to query some inexistent data. If that's the data you just deleted (keyspace / table), this is expected. If not there is an other issue, I hope not related to the delete in this case...

C*heers,
-----------------------
Alain Rodriguez - @arodream - alain@thelastpickle.com
France / Spain

The Last Pickle - Apache Cassandra Consulting



2018-01-12 7:14 GMT+00:00 Eunsu Kim <eunsu.bill23@gmail.com>:
hi everyone

On the development server, I dropped all the tables and even keyspace dropped to change the table schema.
Then I created the keyspace and the table.

However, the actual size of the data directory did not decrease at all. Disk Load monitored by JMX has been decreased.

<Screen Shot 2018-01-12 at 3.50.22 PM.png>


After that, Cassandra does not work normally.

The client fails to establish a connection and I see the following exceptions in the Cassandra logs.

org.apache.cassandra.db.UnknownColumnFamilyException: Couldn't find table for cfId…….org.apache.cassandra.io.FSReadError: java.io.IOException: Digest mismatch exception……


After the data is forcibly deleted, Cassandra is restarted in a clean state and works well.

Can anyone guess why this is happening?

Thank you in advance.