zookeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Flavio Junqueira <...@apache.org>
Subject Re: Zookeeper jmx monitoring for kafka
Date Sat, 12 Sep 2015 14:44:40 GMT
We do support jmx extensively, and I believe you can use it Ganglia. As for metrics, things
that come to mind now on top of what Otis suggested are:

- Frequency of leader election (if they happen too frequently, then that's going to affect
your availability and probably an indication of something wrong)
- Disk space utilization
- Number of sessions open per server
- Number of znodes, database size and such

You can also use four-letter words to get some info out of servers. Check the documentation
for the ones available. 

-Flavio

> On 12 Sep 2015, at 02:25, Otis Gospodnetić <otis.gospodnetic@gmail.com> wrote:
> 
> Hi Prabhjot,
> 
> Short answer: yes
> I used to think ZK was so super stable that it was one of those things that
> don't require any management, but on a few occasions I witnessed complex
> distributed applications nearly fall apart because of issues with ZK.  We
> use our own SPM for ZooKeeper to monitor all our ZK nodes.  I'm not a ZK
> expert, but metrics like Request Latency, Watch Count, Outstanding Requests
> seem important.
> 
> Otis
> --
> Monitoring * Alerting * Anomaly Detection * Centralized Log Management
> Solr & Elasticsearch Support * http://sematext.com/
> 
> 
> On Tue, Aug 18, 2015 at 9:11 AM, Prabhjot Bharaj <prabhbharaj@gmail.com>
> wrote:
> 
>> Hi,
>> 
>> Zookeeper monitoring - I was going through the 'Monitoring' section in
>> http://kafka.apache.org/coding-guide.html
>> 
>> I have setup Ganglia to monitor all the stats inocming from kafka jmx port
>> But, in addition, there are a bunch of stats that Zookeeper also exposes in
>> its jmx port.
>> They are documented here:
>> https://zookeeper.apache.org/doc/r3.3.4/zookeeperJMX.html
>> 
>> Is it worthwhile setting up Ganglia for Zookeeper along with Kafka ?
>> If yes, in your experience, what are the essential stats to look out for
>> (especially in relation to its use with Kafka)
>> 
>> Looking forward to your response.
>> 
>> Thanks,
>> prabcs
>> 


Mime
View raw message