Limit MBeans query to certain scopes #63
Conversation
JMXFetch is currently fetching all available beans regardless of the user configuration. This might cause a high memory load when the application is exposing a lot of metrics (thus OOM issues). This change is the first step towards a bigger JMXFetch memory consumption improvement: the idea is to find, among the user defined beans, some common patterns to limit the scopes to query. For instance, ``` - org.datadog.jmxfetch.test:scope=sameScope,param=sameParam,type=sameType - org.datadog.jmxfetch.test:scope=sameScope,param=notTheSameParam,type=sameType ``` would induce a query on "org.datadog.jmxfetch.test:scope=sameScope,type=sameType,*" instead of "*:*". At the moment, only one scope per domain name is computed: it's kind of the greatest common divisor/pattern of the beans. Ideally, this should be improved to recognize multiple scopes among the same domain name.
| } | ||
| } | ||
| catch (Exception e) { | ||
| e.printStackTrace(); |
There was a problem hiding this comment.
Bit of a nitpick, but could we also log an error so that the exception is not completely silent when JMXFetch's stderr is not redirected to a log?
Something like LOGGER.error("Unable to compute common bean scope, querying all beans as a fallback")
There was a problem hiding this comment.
(not a nitpick and it would be better to catch a more specific exception)
There was a problem hiding this comment.
Good idea.
Otherwise I can't think of a specific exception that i want to catch. I added this very generic try {} catch(){} statement in order to fallback in the previous behavior in case anything was turning wrong. I should maybe remove it completely.
|
Looks great! 👍 Nothing really important in my comments except for the exception thing. |
Log common scope computing exceptions. Improve the code readability.
|
Thanks for the feedback. This closes the JMXFetch v0.8.0 release: I'll write the CHANGELOG and submit the new artifact to dd-agent. |
Limit MBeans query to certain scopes
Do not scope MBeans queries on `list_everything` or `list_not_matching_attributes`. Fix regression introduced with #63.
Do not scope MBeans queries on `list_everything` or `list_not_matching_attributes`. Fix regression introduced with #63.
JMXFetch is currently fetching all available beans regardless of the
user configuration. This might cause a high memory load when the
application is exposing a lot of metrics (thus OOM issues).
This change is the first step towards a bigger JMXFetch memory
consumption improvement: the idea is to find, among the user defined
beans, some common patterns to limit the scopes to query.
For instance,
would induce a query on
org.datadog.jmxfetch.test:scope=sameScope,type=sameType,*instead of
*:*.At the moment, only one scope per domain name is computed: it's kind of
the greatest common divisor/pattern of the beans.
Ideally, this should be improved to recognize multiple scopes among the
same domain name.