You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@activemq.apache.org by Ben Warrick <ho...@gmail.com> on 2020/12/21 17:02:17 UTC

Good monitoring practices, JMX

Hello,

I'm trying to put together a good monitoring strategy to keep Artemis
healthy. I set up Zabbix, which is working pretty well, and I'm using
jconsole to explore the JMX MBeans and troubleshoot.

Right now I only have two alerts. One that watech for the secondary cluster
to come up, and sends an email alert that the primary has failed. And a
second where I divide HeapMemoryUsage.used by HeapMemoryUsage.max and send
an alert if it's greater than 0.7.

It's been suggested that I should monitor garbage collections -- the amount
of CPU time spent and number of collections -- to detect if I'm running out
of memory.

I've also tried out this Zabbix template:
https://github.com/topicusoverheid/zabbix-activmq-template  It doesn't seem
to work in Artemis though. After importing it, two "applications" are
created (in Zabbix terms) but no "items" are created. I've tried out some
of the MBeans it uses though. I don't think this is what I need though.

Is there any guidance on this? A best practices or example showing all the
different MBeans and how they're used to detect the kinds of problems
Artemis is likely to run into?

Thanks,

Ben

Re: Good monitoring practices, JMX

Posted by Justin Bertram <jb...@apache.org>.
FWIW, Zabbix supports grabbing data from Prometheus endpoints [1] as well.


Justin

[1]
https://www.zabbix.com/documentation/4.4/manual/config/items/itemtypes/prometheus

On Mon, Dec 21, 2020 at 2:31 PM Justin Bertram <jb...@apache.org> wrote:

> If you're open to using something other than Zabbix you might consider a
> solution like Prometheus with Grafana. Artemis supports a pluggable metrics
> architecture [1] and there's a Prometheus plugin available here [2]. Both
> Prometheus and Grafana are easy to set up and use and they're both quite
> powerful as well as free. A sample Grafana dashboard is available here [3].
> Aside from relevant broker and queue statistics JVM GC, memory, and thread
> stats are also exported.
>
> Whether you go with a metrics plugin or a more traditional JMX based
> monitoring solution, probably the main thing you're going to want to keep
> an eye on is your message counts. A message broker isn't designed to be a
> data store so when message counts start to increase a lot things are going
> to start behaving differently depending on your address-full-policy
> configuration.
>
> Beyond that I'm not sure what else to recommend. So much depends on your
> use-case which you haven't really described at all.
>
> The Zabbix "template" you cited is designed for ActiveMQ 5.x so it won't
> work with Artemis (which has completely different MBeans).
>
> The JVM you're using should have MBeans for memory, GC, and thread stats.
>
>
> Justin
>
> [1]
> http://activemq.apache.org/components/artemis/documentation/latest/metrics.html
> [2] https://github.com/jbertram/artemis-prometheus-metrics-plugin
> [3] https://github.com/artemiscloud/grafana
>
> On Mon, Dec 21, 2020 at 11:02 AM Ben Warrick <ho...@gmail.com>
> wrote:
>
>> Hello,
>>
>> I'm trying to put together a good monitoring strategy to keep Artemis
>> healthy. I set up Zabbix, which is working pretty well, and I'm using
>> jconsole to explore the JMX MBeans and troubleshoot.
>>
>> Right now I only have two alerts. One that watech for the secondary
>> cluster
>> to come up, and sends an email alert that the primary has failed. And a
>> second where I divide HeapMemoryUsage.used by HeapMemoryUsage.max and send
>> an alert if it's greater than 0.7.
>>
>> It's been suggested that I should monitor garbage collections -- the
>> amount
>> of CPU time spent and number of collections -- to detect if I'm running
>> out
>> of memory.
>>
>> I've also tried out this Zabbix template:
>> https://github.com/topicusoverheid/zabbix-activmq-template  It doesn't
>> seem
>> to work in Artemis though. After importing it, two "applications" are
>> created (in Zabbix terms) but no "items" are created. I've tried out some
>> of the MBeans it uses though. I don't think this is what I need though.
>>
>> Is there any guidance on this? A best practices or example showing all the
>> different MBeans and how they're used to detect the kinds of problems
>> Artemis is likely to run into?
>>
>> Thanks,
>>
>> Ben
>>
>

Re: Good monitoring practices, JMX

Posted by Justin Bertram <jb...@apache.org>.
If you're open to using something other than Zabbix you might consider a
solution like Prometheus with Grafana. Artemis supports a pluggable metrics
architecture [1] and there's a Prometheus plugin available here [2]. Both
Prometheus and Grafana are easy to set up and use and they're both quite
powerful as well as free. A sample Grafana dashboard is available here [3].
Aside from relevant broker and queue statistics JVM GC, memory, and thread
stats are also exported.

Whether you go with a metrics plugin or a more traditional JMX based
monitoring solution, probably the main thing you're going to want to keep
an eye on is your message counts. A message broker isn't designed to be a
data store so when message counts start to increase a lot things are going
to start behaving differently depending on your address-full-policy
configuration.

Beyond that I'm not sure what else to recommend. So much depends on your
use-case which you haven't really described at all.

The Zabbix "template" you cited is designed for ActiveMQ 5.x so it won't
work with Artemis (which has completely different MBeans).

The JVM you're using should have MBeans for memory, GC, and thread stats.


Justin

[1]
http://activemq.apache.org/components/artemis/documentation/latest/metrics.html
[2] https://github.com/jbertram/artemis-prometheus-metrics-plugin
[3] https://github.com/artemiscloud/grafana

On Mon, Dec 21, 2020 at 11:02 AM Ben Warrick <ho...@gmail.com>
wrote:

> Hello,
>
> I'm trying to put together a good monitoring strategy to keep Artemis
> healthy. I set up Zabbix, which is working pretty well, and I'm using
> jconsole to explore the JMX MBeans and troubleshoot.
>
> Right now I only have two alerts. One that watech for the secondary cluster
> to come up, and sends an email alert that the primary has failed. And a
> second where I divide HeapMemoryUsage.used by HeapMemoryUsage.max and send
> an alert if it's greater than 0.7.
>
> It's been suggested that I should monitor garbage collections -- the amount
> of CPU time spent and number of collections -- to detect if I'm running out
> of memory.
>
> I've also tried out this Zabbix template:
> https://github.com/topicusoverheid/zabbix-activmq-template  It doesn't
> seem
> to work in Artemis though. After importing it, two "applications" are
> created (in Zabbix terms) but no "items" are created. I've tried out some
> of the MBeans it uses though. I don't think this is what I need though.
>
> Is there any guidance on this? A best practices or example showing all the
> different MBeans and how they're used to detect the kinds of problems
> Artemis is likely to run into?
>
> Thanks,
>
> Ben
>