Enhancement or can javamelody already send alerts.

[Robert B]

I haven't seen in any of the documentation so I am wondering if
Javamelody has the ability to send an alert on certain thresholds that
are met?

This would be awesome if it could say send alert if something starts
running away with memory usage or something in your database pool just
run crazy and never let goes of a thread.

I know we can get daily, weekly and monthly emails for trending and
such but it would be great as a Application Server Admin to be able to
set alerts to notify myself or someone if I don't happen to be
watching the javamelody page at a given time.

Thanks,

Robert

[Darren Hartford]

I myself (as a user) have been struggling with this question.  Javamelody really isn't targetted for this as much as analysis/statistics collection, but with it so easy to install and setup it would be nice for it to also be an SLA-monitoring/notifier tool...but not sure if this is the right place to put it(and possibly reduce focus on it's core competency). monitoring/notification (based on SLA's) can reach deep into the realm of SNMP, SMS, JMX, and various monitoring tools already in this place (jopr, nagios, opennms, jamon, etc for open source ones). FLIP-side...most of those monitoring capabilities/existing tools focus on enterprise monitoring of full systems and not necessarily a specific application, which JavaMelody does well. -D --- On Mon, 7/26/10, Robert B <breaul...@gmail.com> wrote:

[evernat]

Hi Robert,

Sorry, there is no ability to send alerts on thresholds.

In my humble opinion, the big problem with alerts is that a lot are
sent and nobody really reads them.
Alerts and thresholds should be chosen wisely in order not to be
useless.

But sending alerts is on the todo list. So you can check back at a
later date.

bye, Emeric

[evernat]

Hi,

Yes alerts are on the todo list, but not immediately.

And there is already a basic SLA monitoring in javamelody: the warning-
threshold-millis and severe-threshold-millis parameters. These
parameters in milliseconds could be used to display SLA values in
summaries of http requests or sql requests. For example, you could
have global severe percentage (could be "bad" if more than 10% mean
times are higher than the value 20s) and global warning percentage of
http requests mean times. In details of http requests, related colors
are used for mean times: red for severe level, yellow for warning
level and green otherwise.

And I would like to have documentation on configuration of nagios,
munin or others for alerts and SLA.
A Munin (and possibly Nagios) plugin is on the todo list. An
undocumented feature is already available for this: if you have
graphics for usedMemory and cpu available at
http://host:port/context/monitoring?width=200&height=50&graph=usedMemory
You can simply have last values in the graphics with
http://host:port/context/monitoring?part=lastValue&graph=usedMemory,cpu
So a Munin plugin could be easy for knowledgeable people, for example
with something like:
http://github.com/coderholic/munin-popularity-plugins/tree/09d3c45b2bfc196a35c13e80aacd5b33db0eb73d

bye, Emeric

On 28 juil, 02:39, Darren Hartford <binarymon...@yahoo.com> wrote:
> I myself (as a user) have been struggling with this question.  Javamelody really isn't targetted for this as much as analysis/statistics collection, but with it so easy to install and setup it would be nice for it to also be an SLA-monitoring/notifier tool...but not sure if this is the right place to put it(and possibly reduce focus on it's core competency).
> monitoring/notification (based on SLA's) can reach deep into the realm of SNMP, SMS, JMX, and various monitoring tools already in this place (jopr, nagios, opennms, jamon, etc for open source ones).
> FLIP-side...most of those monitoring capabilities/existing tools focus on enterprise monitoring of full systems and not necessarily a specific application, which JavaMelody does well.
> -D
>

> --- On Mon, 7/26/10, Robert B <breaultrob...@gmail.com> wrote:

[LightGuard]

On Jul 28, 8:30 am, evernat <ever...@free.fr> wrote:
> Hi Robert,
>
> Sorry, there is no ability to send alerts on thresholds.
>
> In my humble opinion, the big problem with alerts is that a lot are
> sent and nobody really reads them.
> Alerts and thresholds should be chosen wisely in order not to be
> useless.
>
> But sending alerts is on the todo list. So you can check back at a
> later date.

Would it make more sense to create an SPI for this and expose hooks so
that other modules could be built upon Java Melody that could provide
this functionality?

[evernat]

Hi LightGuard,

There is already an SPI for this in javamelody:
http://host:port/context/monitoring?part=lastValue&graph=usedMemory,cpu
as said in the other post of this thread (usedMemory and cpu are only
examples).

This SPI seems good for Munin and Nagios plugins. Or what would be
needed for such a plugin?

bye, Emeric