Suggestions for PIQL processor/process resource/alert engine

[Alan Evans]

Excuse my recent SPAM.  I really dig RHQ, I think it has a lot of
potential but there are few things I would like it to do.  Ok on to my
suggestions.

PIQL:
-- A way to select a process based on user, and group attributes?
user, effective user, login user, group, effective group, login group
and by name or uid/gid?
  maybe:
    process|user|foo
    process|uid|500
    process|group|foo
    process|gid|500
    process|euser|bar
    process|luid|500

Process resource:
-- A way to return a number of processes matching a PIQL as a trait?
Or maybe a trait for number of processes that match?  This way an
alert could be made for num process > = < some number.  I.e. we have a
check that looks to make sure there are not more than a certain number
of java processes running.  Otherwise it indicates that batch jobs are
not completing.
-- What about returning information that could be retrieved using
uname as traits like max open file descriptors, max processes etc

Alert Engine:
-- The ability to compare one attribute to another or an attribute to
a trait?  Or a percentage of trait?
  Alert when open files > 90% max open files
  Alert when Java Heap > 90% max heap size
-- Perhaps a way of defining recovery thresholds without having to
define recovery alerts.  Maybe wizard to build a recovery alert that
is created under the hood but hidden from the alert definition view.
Maybe like having the conditions tab have a "recovery conditions" tab
and "recovery damping" or recovery tab with "conditions" and "dapming"
subtabs?  It is cumbersome to have to create a recovery alert and then
its associated alert, or create an alert then its recovery alert then
go edit the original again to assign its recovery alert. (If that
makes sense...)

I would love to hear other people's thoughts and whether or not these
ideas might be useful others.

Regards,
-Alan

[Deon Lackey]

I am not an expert on this, but I think some of the resource tasks you 
want can be done through the RHQ CLI[0]. Like creating a new resource[1] 
and getting a list of resources by trait[2]. (I think. I emphasize -- I 
am not a developer. I am just learning all this stuff, too.)

The first alerting scenario you have can already be accomplished through 
the UI. Most resources have a lot of metrics defined already[3]. The 
metrics which are collected are defined in the agent plug-in, in the 
META-INF/rhq-plugin.xml file. If you needed some metrics for a resource 
which aren't collected, you may be able to edit the resource plug-in. 
Likewise, if you want a resource that isn't currently in RHQ, you can 
create your own plug-in. There's a generator you can use to build a 
skeleton plugin, with details at 
http://rhq-project.org/display/RHQ/Plugins+-+Plugin+Generator.

As for the recovery alerts -- yeah, I agree. It would be really nice to 
define the recovery alert in the alert definition itself.

Cheers,
Deon

[0] http://rhq-project.org/display/JOPR2/Running+the+RHQ+CLI

[1] 
http://docs.redhat.com/docs/en-US/JBoss_Operations_Network/100/html/API/remo...

[2] 
http://docs.redhat.com/docs/en-US/JBoss_Operations_Network/100/html/API/remo...

[3] http://rhq-project.org/display/JOPR2/Managed+Resources

On 12/12/2011 6:51 PM, Alan Evans wrote:
> Excuse my recent SPAM.  I really dig RHQ, I think it has a lot of
> potential but there are few things I would like it to do.  Ok on to my
> suggestions.
>
> PIQL:
> -- A way to select a process based on user, and group attributes?
> user, effective user, login user, group, effective group, login group
> and by name or uid/gid?
>    maybe:
>      process|user|foo
>      process|uid|500
>      process|group|foo
>      process|gid|500
>      process|euser|bar
>      process|luid|500
>
> Process resource:
> -- A way to return a number of processes matching a PIQL as a trait?
> Or maybe a trait for number of processes that match?  This way an
> alert could be made for num process>  =<  some number.  I.e. we have a
> check that looks to make sure there are not more than a certain number
> of java processes running.  Otherwise it indicates that batch jobs are
> not completing.
> -- What about returning information that could be retrieved using
> uname as traits like max open file descriptors, max processes etc
>
> Alert Engine:
> -- The ability to compare one attribute to another or an attribute to
> a trait?  Or a percentage of trait?
>    Alert when open files>  90% max open files
>    Alert when Java Heap>  90% max heap size
> -- Perhaps a way of defining recovery thresholds without having to
> define recovery alerts.  Maybe wizard to build a recovery alert that
> is created under the hood but hidden from the alert definition view.
> Maybe like having the conditions tab have a "recovery conditions" tab
> and "recovery damping" or recovery tab with "conditions" and "dapming"
> subtabs?  It is cumbersome to have to create a recovery alert and then
> its associated alert, or create an alert then its recovery alert then
> go edit the original again to assign its recovery alert. (If that
> makes sense...)
>
> I would love to hear other people's thoughts and whether or not these
> ideas might be useful others.
>
> Regards,
> -Alan
> _______________________________________________
> rhq-users mailing list
> [e-mail]
> https://fedorahosted.org/mailman/listinfo/rhq-users