problem installing a custom JMX plugin when the defaultInterval is < 30000

[BARTH Michael]

Hello All,

I am trying to get some JMX metrics into RHQ. All works fine but I have some trouble
installing the JMX plugin whith a defaultInterval lower than 30000 (a
custom-jmx-plugin-1.0-SNAPSHOT.jar.fixme is left in the update directory when clicking on
"scan for updates" in RHQ's Administration tab).

A JMX plugin with the following metrics inside deploys:

        <metric property="ServiceTime"
              displayName="Service time"
              defaultOn="true"
              displayType="summary"
              defaultInterval="30000"
              description="Time between the first and request and the last
response."/>

        <metric property="ServiceTime"
              displayName="Service time"
              defaultOn="true"
              displayType="summary"
              defaultInterval="40000"
              description="Time between the first and request and the last
response."/>

A JMX plugin with the following metrics inside does not deploy:

        <metric property="ServiceTime"
              displayName="Service time"
              defaultOn="true"
              displayType="summary"
              defaultInterval="20000"
              description="Time between the first and request and the last
response."/>


Any help on this would be highly appreciated.
Thank, Michael

[John Mazzitelli]

On 06/09/2011 05:27 AM, BARTH Michael wrote:
>  some trouble installing the JMX plugin whith a defaultInterval lower
>  than 30000 (a custom-jmx-plugin-1.0-SNAPSHOT.jar.fixme is left in the
>  update directory when clicking on "scan for updates" in RHQ's
>  Administration tab).
>  
As per the documentation/annotation found in the XML Schema for plugin 
descriptors (rhq-plugin.xsd):

http://git.fedorahosted.org/git/?p=rhq/rhq.git;a=blob;f=modules/core/clie...

default collection intervals must be 30s or higher. See line 201-214 of 
rhq-plugin.xsd where it defines the defaultInterval schema:

201     <xs:attribute name="defaultInterval">
202        <xs:annotation>
203           <xs:documentation>
204              The default collection interval, in milliseconds. The 
metric is not necessarily
205              enabled - see the defaultOn attribute for that. The 
minimum allowed value is
206              30000 (30 seconds).
207           </xs:documentation>
208        </xs:annotation>
209        <xs:simpleType>
210           <xs:restriction base="xs:unsignedLong">
211              <xs:minInclusive value="30000"/>
212           </xs:restriction>
213        </xs:simpleType>
214     </xs:attribute>

So, had you been using an XML Editor that validates your plugin 
descriptor against this XSD schema, it should have flagged that as an 
invalid value.

We shouldn't silently ignore this though - if we aren't logging an error 
message at plugin deployment time, we should be. It would be a problem 
if we do not log an error to let you know why it failed.

>  A JMX plugin with the following metrics inside deploys:
>
>  <metric property="ServiceTime"
>  displayName="Service time"
>  defaultOn="true"
>  displayType="summary"
>  defaultInterval="30000"
>  description="Time between the first and request and the last response."/>
>
>  <metric property="ServiceTime"
>  displayName="Service time"
>  defaultOn="true"
>  displayType="summary"
>  defaultInterval="40000"
>  description="Time between the first and request and the last response."/>
>
>  A JMX plugin with the following metrics inside does not deploy:
>
>  <metric property="ServiceTime"
>  displayName="Service time"
>  defaultOn="true"
>  displayType="summary"
>  defaultInterval="20000"
>  description="Time between the first and request and the last response."/>
>
>
>  Any help on this would be highly appreciated.
>  Thank, Michael
>
>
>
>  _______________________________________________
>  rhq-users mailing list
>  rhq-users(a)lists.fedorahosted.org
>  https://fedorahosted.org/mailman/listinfo/rhq-users

[Ian Springer]

>  As per the documentation/annotation found in the XML Schema for
> plugin
>  descriptors (rhq-plugin.xsd):
>
>
>  default collection intervals must be 30s or higher. See line 201-214 of
>  rhq-plugin.xsd where it defines the defaultInterval schema:
>
>
>  So, had you been using an XML Editor that validates your plugin
>  descriptor against this XSD schema, it should have flagged that as an
>  invalid value.
>
>  We shouldn't silently ignore this though - if we aren't logging an error
>  message at plugin deployment time, we should be. It would be a problem
>  if we do not log an error to let you know why it failed. 
I just tried deploying a plugin with one of the metric's default 
collection interval set to 20000. The following warning was printed in 
the Server log:

08:58:22,067 WARN  [AgentPluginScanner] Failed to scan agent plugin 
[/home/ips/Projects/rhq/dev-container/jbossas/server/default/deploy/rhq.ear/rhq-downloads/rhq-plugins/jopr-jboss-as-plugin-4.1.0-SNAPSHOT.jar]

found on filesystem. Skipping. Cause: 
org.rhq.core.clientapi.agent.PluginContainerException: Could not 
successfully parse the plugin descriptor [META-INF/rhq-plugin.xml found 
in plugin jar at 
[file:/home/ips/Projects/rhq/dev-container/jbossas/server/default/deploy/rhq.ear/rhq-downloads/rhq-plugins/jopr-jboss-as-plugin-4.1.0-SNAPSHOT.jar]

It correctly states that it could not parse the plugin descriptor, but 
it doesn't give any details as to what was wrong with the plugin 
descriptor. We can probably print something more informative here. I'll 
take a peek at the code.

[Jay Shaughnessy]

The 30s minimum is to protect against metric flooding and ensure that 
the plugin can report on metric requests in a fast enough manner.

On 6/10/2011 9:04 AM, Ian Springer wrote:
> > As per the documentation/annotation found in the XML Schema for
> plugin
> > descriptors (rhq-plugin.xsd):
> >
> >
> > default collection intervals must be 30s or higher. See line 201-214 of
> > rhq-plugin.xsd where it defines the defaultInterval schema:
> >
> >
> > So, had you been using an XML Editor that validates your plugin
> > descriptor against this XSD schema, it should have flagged that as an
> > invalid value.
> >
> > We shouldn't silently ignore this though - if we aren't logging an error
> > message at plugin deployment time, we should be. It would be a problem
> > if we do not log an error to let you know why it failed.
>  I just tried deploying a plugin with one of the metric's default
>  collection interval set to 20000. The following warning was printed in
>  the Server log:
>
>  08:58:22,067 WARN  [AgentPluginScanner] Failed to scan agent plugin
>
> [/home/ips/Projects/rhq/dev-container/jbossas/server/default/deploy/rhq.ear/rhq-downloads/rhq-plugins/jopr-jboss-as-plugin-4.1.0-SNAPSHOT.jar]
>  found on filesystem. Skipping. Cause:
>  org.rhq.core.clientapi.agent.PluginContainerException: Could not
>  successfully parse the plugin descriptor [META-INF/rhq-plugin.xml found
>  in plugin jar at
>
> [file:/home/ips/Projects/rhq/dev-container/jbossas/server/default/deploy/rhq.ear/rhq-downloads/rhq-plugins/jopr-jboss-as-plugin-4.1.0-SNAPSHOT.jar]
>
>  It correctly states that it could not parse the plugin descriptor, but
>  it doesn't give any details as to what was wrong with the plugin
>  descriptor. We can probably print something more informative here. I'll
>  take a peek at the code.
>
>  _______________________________________________
>  rhq-users mailing list
>  rhq-users(a)lists.fedorahosted.org
>  https://fedorahosted.org/mailman/listinfo/rhq-users
>

[Ian Springer]

We shouldn't silently ignore this though - if we aren't logging an error
> > message at plugin deployment time, we should be. It would be a
> problem
> > if we do not log an error to let you know why it failed.
>  I just tried deploying a plugin with one of the metric's default
>  collection interval set to 20000. The following warning was printed in
>  the Server log:
>
>  08:58:22,067 WARN  [AgentPluginScanner] Failed to scan agent plugin
>
> [/home/ips/Projects/rhq/dev-container/jbossas/server/default/deploy/rhq.ear/rhq-downloads/rhq-plugins/jopr-jboss-as-plugin-4.1.0-SNAPSHOT.jar]
>  found on filesystem. Skipping. Cause:
>  org.rhq.core.clientapi.agent.PluginContainerException: Could not
>  successfully parse the plugin descriptor [META-INF/rhq-plugin.xml found
>  in plugin jar at
>
> [file:/home/ips/Projects/rhq/dev-container/jbossas/server/default/deploy/rhq.ear/rhq-downloads/rhq-plugins/jopr-jboss-as-plugin-4.1.0-SNAPSHOT.jar]
>
>  It correctly states that it could not parse the plugin descriptor, but
>  it doesn't give any details as to what was wrong with the plugin
>  descriptor. We can probably print something more informative here. I'll
>  take a peek at the code. 

I checked in a fix to master [1] so that we now log the details of the 
validation errors, including the line # and column # of the schema 
violation. Here's an example:

[ERROR] Validation fatal error while parsing
[jopr-jboss-as-plugin-4.1.0-SNAPSHOT.jar:META-INF/rhq-plugin.xml] at line 221,
column 94: cvc-minInclusive-valid: Value '20000' is not facet-valid with
respect to minInclusive '30000' for type
'#AnonType_defaultIntervalmetric'.

Similar logging is also done when validation errors occur during parsing 
of server plugin descriptors.

[1] https://bugzilla.redhat.com/show_bug.cgi?id=712444