Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • 09:18
    ivantopo opened #1089
  • Nov 24 15:57
    MaciejSzewczyszyn closed #1077
  • Nov 24 15:57
    MaciejSzewczyszyn commented #1077
  • Nov 24 15:56
    MaciejSzewczyszyn closed #1078
  • Nov 23 15:08
    getArtemUsername commented #1075
  • Nov 23 14:58

    ivantopo on v2.4.1

    (compare)

  • Nov 23 14:54

    ivantopo on master

    Add jvm threads states metrics … (compare)

  • Nov 23 14:54
    ivantopo closed #1075
  • Nov 23 14:52
    ivantopo commented #1075
  • Nov 23 14:46
    getArtemUsername commented #1075
  • Nov 23 14:13

    ivantopo on v2.4.0

    (compare)

  • Nov 22 09:25

    ivantopo on master

    reduce default retries and back… proactively drop old spans in t… (compare)

  • Nov 18 09:25

    ivantopo on master

    fix thread pool type on schedul… (compare)

  • Nov 18 07:40
    getArtemUsername synchronize #1075
  • Nov 17 17:33
    getArtemUsername synchronize #1075
  • Nov 17 17:32
    getArtemUsername synchronize #1075
  • Nov 17 15:59
    ivantopo commented #1077
  • Nov 17 15:55
    MaciejSzewczyszyn commented #1077
  • Nov 17 15:54
    ivantopo edited #1082
  • Nov 17 15:53
    ivantopo opened #1082
Ivan Topolnjak
@ivantopo
that would happen when the 99% has the same value as the 87.5% (I think), I'm reasoning about it now
yeah, regardless of whether a few percentiles actually have the same value we should return the same p that was passed in as a parameter
@moriyasror when you say that you only get the metric once at start, what exactly do you mean? that metric is only recorded when a connection is closed, could it be that you get persistent connections that actually last for really long times?
John Watson
@jkwatson
@ivantopo should I log a bug about a different p being returned that is passed in?
Ivan Topolnjak
@ivantopo
yeap
I will most likely fix it later today
John Watson
@jkwatson
ok, will do! thanks!
Ivan Topolnjak
@ivantopo
:thumbsup:
John Watson
@jkwatson
Ivan Topolnjak
@ivantopo
thanks!
moriyasror
@moriyasror
@ivantopo I run my service locally via intellij and send several http requests via postman.
I see the 'http.server.connection.lifetime' metric only once, even though I see other metrics like http.server.requests
moriyasror
@moriyasror
Hi, I mange to run kamon from my intellij, but failed to do it on a server because of missing JDK.
My question is why I must have JDK? on the production we have only JRE, the JDK is only from development.
Is there a way to run kamon using JRE only?
Ivan Topolnjak
@ivantopo
yes, you will have to use the -javaagent way of setting up the agent instead of Kamon.init()
what kind of application were you trying to instrument?
moriyasror
@moriyasror
@ivantopo I instrument akka-http.
can you explain more about the -javaagent option?
Ivan Topolnjak
@ivantopo
sure.. so, I'm guessing that what you did right now was to add the kamon-bundle and call Kamon.init() when application starts, correct?
moriyasror
@moriyasror
@ivantopo Generally yes, I also change the configuration to get only akka-http metrics, and warp the Kamon.init() to run base on configuration (to be able to disable it in case of issue)
My code is written in scala, I saw some plugin which can add the agent, I will try to use them.
If I remove the Kamon.init(), from my code, will I be able to control the kamon running and disable it?
Ivan Topolnjak
@ivantopo
you will still need to have the Kamon.init call in your code because that's what starts all the reporters, status page and so on
the best way to add the -javaagent option would be using the sbt-javaagent plugin, we mention it at the end of this page: https://kamon.io/docs/latest/guides/installation/setting-up-the-agent/
when you start your application with the agent, by the time Kamon.init is executed the agent is already there so there is no need to attach anything in runtime and no need to have a JDK
moriyasror
@moriyasror
@ivantopo thanks! I mange to run it via the plugin.
Guillaume Noireaux
@gnoireaux

Hello,

What is the recommended way to propagate context manually?
I'd like to follow a process through multiple services (Play Scala). They exchange messages via rabbitmq.
Would it be as simple as using AMQP headers the same way HTTP headers are used? How to do that?

Can I use something like https://github.com/opentracing-contrib/java-rabbitmq-client with Kamon and then Jaeger export?

Thanks for your patience!

5 replies
armelouche
@armelouche
Hello, (thank you for you work first), are there any basic Grafana dashboards available for Kamon 2 like there was Kamon 1 (jvm..) ?
Ivan Topolnjak
@ivantopo
@armelouche we don't have any "official" dashboards but there are a few created by folks around the globe on the Grafana site: https://grafana.com/grafana/dashboards?direction=asc&orderBy=name&search=kamon
1 reply
Sergey Morgunov
@ihostage
@ivantopo Please, any reaction to kamon-io/Kamon#760 :pray:
Tell me please, what I need to do for this PR will be merged? :joy:
Ivan Topolnjak
@ivantopo
hey man
:sob:
Sergey Morgunov
@ihostage
:wave:
Ivan Topolnjak
@ivantopo
things have been really crazy on the other side of the job
and man, every single time I promise to check something with a time in mind I fail, the best intentions not always align with what the world has planned for me :joy:
Sergey Morgunov
@ihostage

things have been really crazy on the other side of the job

Me too :joy: But I believe, that sooner or later we can do that together :joy:

Ivan Topolnjak
@ivantopo
yeap, I really appreciate your persistence man, it is really motivating to see folks who care!
Sergey Morgunov
@ihostage
:+1: :wink:
Time to time I will be ping you and I will not fall behind. :joy:
Ivan Topolnjak
@ivantopo
:joy: thanks!
Arsene
@Tochemey
@ihostage quick one is your PR able to help instrument a lagom base application? We had some issue when doing instrumentation in lagom using kamon.
Sergey Morgunov
@ihostage
@Tochemey It's just the first step — implementation of Lagom Circuit Breakers Metrics SPI.
Still, it not fixed a problem with using Kamon in Lagom DEV-Mode. I will try to fix that in the future when I find time for that :smile:
Arsene
@Tochemey
@ihostage :smiley:
Ivan Topolnjak
@ivantopo
the issue of lagom in dev mode is way more complicated because in that case there are (potentially) several different services running on the same JVM
together with SBT and all its stuff
:/
currently Kanela applies all instrumentation on the entire JVM and yeah, we have some filters to prevent touching some classloaders
but that is not enough to account for the fact that some of the services might have Kamon on their classpaths and some others not
plus, Kamon itself doesn't allow for several services on the same JVM
in fact, there has never been a need for it, only for Lagom in dev mode
Sergey Morgunov
@ihostage
Yep :smile:
Ghost
@ghost~5cc594f1d73408ce4fbee018
Hi @ivantopo, for the SQS queue, I’ve opted to manually create a span and pass within the objects that I pass down stream and do something as follows
private val deserializeAndProcessStage = Flow[SpanContext[SqsQueueMessage]]
    .via(killSwitch.flow)
    .throttle(throttleTotal, 1.second, throttleTotal, ThrottleMode.shaping)
    .mapAsync(parallelism) {
      case SpanContext(span, SqsQueueMessage(q, message)) =>
        Kamon.runWithSpan(span) {
          val parsedMsg = ArchonPayload.parse(message)
          logger.debugWithData(s"Message parsed", Map("message" -> parsedMsg.toString))
          parsedMsg match {
            case Right(mdsolUri) =>
              processMessage(mdsolUri.resourceUri)
                .map(Right(_))
                .recover {
                  case NonFatal(th) =>
                    Left(th)
                }
                .map(res => SpanContext(span, SqsQueueMessageResult(q, message, res)))

            case Left(parsingError) =>
              logger.errorWithData(s"An Error has occurred while ", parsingError.th, Map.empty)
              Future.successful(SpanContext(span, SqsQueueMessageResult(q, message, Right(parsingError))))
          }
        }
    }
and as for the span that’s been waiting for 17 days, it happens on my local machine with a new start of the app connecting to a locally run open-zipkin docker machine
Zvi Mints
@ZviMints

Hey all, i have a problem with Kamon and Datadog
This is my configurations:

# Monitoring
kamon {
  datadog {
    flush-interval = 10 seconds
    hostname = datadog
    port = 8125
    application-name = ${?app.name}
    time-units = ms
    memory-units = mb

    subscriptions {
          system-metric = [ ]
          http-server = [ ]

    }
  }
  metric {
    tick-interval = 10 seconds
    track-unmatched-entities = no
    filters.trace.includes = []
  }
}

I connected to some Amazon Queues and Read messages from there, for each message that i insert to database I'm do:

Kamon.metrics.counter(s"message_pulled", Map("queueName" -> queueName)).increment()

BUT I'm getting that i have 100 messages in database for 12:20 for example but in datadog logs i have only 3, how its possible?
Thanks!

Paul Bernet
@pbernet
Can I expose data via JMX with the new "kamon-bundle" "2.1.0"