How does nodetool captures all the metrics about cassandra node

[Kant Kodali]

How does nodetool capture all the metrics about cassandra node? Does cassandra writes all the metrics to some file and nodetool(daemon) will just read off from it?

[Bo Finnerup Madsen]

Hi Kant,

Nodetool uses cassandras JMX interface to fetch metrics. You can read about the JMX interface here:

http://docs.datastax.com/en/cassandra/2.0/cassandra/operations/ops_monitoring_c.html?scroll=concept_ds_pkn_drq_ck__opsMonitoringJconsole

https://wiki.apache.org/cassandra/JmxInterface

https://www.datadoghq.com/blog/how-to-collect-cassandra-metrics/

tor. 5. maj 2016 kl. 23.31 skrev Kant Kodali <ka...@peernova.com>:

How does nodetool capture all the metrics about cassandra node? Does cassandra writes all the metrics to some file and nodetool(daemon) will just read off from it?

--
You received this message because you are subscribed to the Google Groups "DataStax Java Driver for Apache Cassandra User Mailing List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to java-driver-us...@lists.datastax.com.

[Kant Kodali]

Hi,

I am still not understanding how these beans work (perhaps I need to take a deep dive). It will be a great help if someone can point out if the metrics are written to a log or table and a daemon reads off from one of them?

Thanks!

[Andrew Tolbert]

Hi Kant,

I am still not understanding how these beans work (perhaps I need to take a deep dive). It will be a great help if someone can point out if the metrics are written to a log or table and a daemon reads off from one of them?

The metrics themselves aren't written anywhere (unless you set up a metrics reporter).  They are all recorded in memory and exposed as JMX beans.   The nodetool command makes the appropriate JMX calls to grab those bean attribute values.  The code for nodetool is in the org.apache.cassandra.tools.nodetool package with Nodetool.java being the point of entry.   For example, the cfstats metrics are accessed via TableStats where you can see it makes the appropriate calls to the mbeans to get the values via NodeProbe.  If you keep code diving, you'll find the origin of the metrics and how they are calculated.

Thanks!

Andy

[Kant Kodali]

Hi Andrew,

Isn't it a lot to keep it in memory? or is it a fixed size? if it doesnt write any metrics to a log file or anything? What happens when the node dies and comes back up I can still use nodetool netstats or nodetool cfstats and be able to get back the the right numbers right?

Thanks,

Kant

[Kant Kodali]

Hi Andrew,

CFStats reports things like Total space occupied by the table, Estimated number of keys in the table and so on. so if cassandra node dies and comes back up does the nodetool cfstats for say Estimated number of keys will be reported zero? since the metrics are not persisted anywhere?

Thanks,

Kant

On Fri, May 6, 2016 at 8:01 AM, Andrew Tolbert <andrew....@datastax.com> wrote:

[Olivier Michallat]

Hi Kant,

On a side note, these questions are not related to the Java driver. The Cassandra users mailing list would be a better place to ask them.

Thanks,

--

Olivier Michallat

Driver & tools engineer, DataStax