| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The new ModuleSupervisor implementation will make use of
special configuration keys under the `kamon.modules` section
to detect all the available modules in the classpath and do
two things with that info:
1. Log the AspectJ Weaver missing error if any of the
available modules requires AspectJ.
2. Start all available modules that are marked with the
auto-start setting.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
module, related to #136.
All Akka-related instrumentation and code has been moved to the kamon-akka module,
including the filters for actor, dispatcher and router metrics. Also the following
changes are included:
- Router Metrics are now working properly, related to #139.
- Cleanup the log output for this module, related to #142.
- Some minor cleanups in various tests.
This PR breaks the reporting modules which will need to wait for #141 to be ready to come back
to life.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
automatically named traces, closes #35
After a bit of testing it seems like instead of generating more friendly names by default, we need to do
changes that are specific to each module (StatsD and Datadog) in order to make it work properly. Here is
a list of the introduced changes:
- StatsD
- Change dots (.) to underscores (_) on the host name string.
- Change ': ' to single underscore and spaced to underscore. Userful for automatically generated http
trace names.
- Change '/' to underscore. Ideally the dashboard could replace '_' with '/' to display it like a path.
- Datadog
- Change ': ' to ':'. Userful for automatically generated http trace names.
|
|
|
|
| |
kamon.metrics.default-collection-context-buffer-size setting
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This PR is including several changes to the kamon-core, most notably:
- Formalize the interface for Histograms, Counters and MinMaxCounters. Making sure
that the interfaces are as clean as possible.
- Move away from the all Vector[Measurement] based Histogram snapshot to a new approach
in which we use a single long to store both the index in the counts array and the
frequency on that bucket. The leftmost 2 bytes of each long are used for storing the
counts array index and the remaining 6 bytes are used for the actual count, and
everything is put into a simple long array. This way only the buckets that actually
have values will be included in the snapshot with the smallest possible memory
footprint.
- Introduce Gauges.
- Reorganize the instrumentation for Akka and Scala and rewrite most of the tests
of this components to avoid going through the subscription protocol to test.
- Introduce trace tests and fixes on various tests.
- Necessary changes on new relic, datadog and statsd modules to compile with the new
codebase.
Pending:
- Finish the upgrade of the new relic to the current model.
- Introduce proper limit checks for histograms to ensure that we never pass the 2/6 bytes
limits.
- More testing, more testing, more testing.
- Create the KamonStandalone module.
|
|
|
|
| |
TraceContext through the filters and all actions in the incoming request
|
| |
|
|
|
|
| |
* I've changed the way to get the Actorsystem inside of DispatcerTracing, passing the actorSystem across the Dispatchers and then to each Dispatcher in the same ActorSystem.
|
| |
|
| |
|
|
|
|
| |
StatsD
|
| |
|
| |
|
| |
|
| |
|
|
|