DataDog and StatsD Metrics Support #1701

aantono · 2017-06-02T02:41:17Z

While still waiting for #1504 to be merged and rebase on top, this is to get the review process for #834 started for DataDog and StatsD metrics support.

The solution is based on the same go-kit/metrics library as for Prometheus, except using the DataDog and StatsD providers. Given that there is a possible integration from DataDog being created by DataDog/integrations-extras#64, I've used the matching metric names to push the metrics to DD agent, as the Prometheus pull is mapping. Internally the DD agent puller is still pushing the collected data into DogStatsD listener, so it will end up flowing though the same ingestion gate, thus taking advantage of the "included integration metrics".

I've had to add the additional dependencies into Glide, so after running ./scripts/glide.sh add <> it has changed a bunch of files in the vendor, so hopefully that's done correctly.

aantono · 2017-06-07T16:31:48Z

Rebased on top of #1504 changes from Master.

ldez · 2017-06-08T08:39:11Z

middlewares/datadog_test.go

+	udp.ShouldReceiveAll(t, expected, func() {
+		n.ServeHTTP(recorder, req1)
+		n.ServeHTTP(recorder, req2)
+		//		body := recorder.Body.String()


ldez · 2017-06-08T08:40:18Z

middlewares/datadog_test.go

+	})
+	n.UseHandler(r)
+
+	req1, err := http.NewRequest("GET", "http://localhost:3000/ok", nil)


req1 := testhelpers.MustNewRequest(http.methodGet, "http://localhost:3000/ok", nil)

ldez · 2017-06-08T08:40:35Z

middlewares/datadog_test.go

+	if err != nil {
+		t.Error(err)
+	}
+	req2, err := http.NewRequest("GET", "http://localhost:3000/not-found", nil)


req2 := testhelpers.MustNewRequest(http.methodGet, "http://localhost:3000/not-found", nil)

ldez · 2017-06-08T08:41:38Z

middlewares/datadog_test.go

+
+	n.Use(metricsMiddlewareBackend)
+	r := http.NewServeMux()
+	//r.Handle("/metrics", promhttp.Handler())


Dead Code ?

ldez · 2017-06-08T08:43:44Z

middlewares/datadog.go

+		}
+		pushInterval, err := time.ParseDuration(config.PushInterval)
+		if err != nil {
+			pushInterval = 10 * time.Second


maybe you can log the error.

ldez · 2017-06-08T08:48:25Z

middlewares/statsd_test.go

+	if err != nil {
+		t.Error(err)
+	}
+	req2, err := http.NewRequest("GET", "http://localhost:3000/not-found", nil)


req2 := testhelpers.MustNewRequest(http.methodGet, "http://localhost:3000/not-found", nil)

ldez · 2017-06-08T08:48:35Z

middlewares/statsd_test.go

+	udp.ShouldReceiveAll(t, expected, func() {
+		n.ServeHTTP(recorder, req1)
+		n.ServeHTTP(recorder, req2)
+		//		body := recorder.Body.String()


Dead code ?

ldez · 2017-06-09T09:27:16Z

glide.lock

  subpackages:
  - log
  - swagger
 - name: github.com/fatih/color
  version: 9131ab34cf20d2f6d83fdc67168a5430d1c7dc23
 - name: github.com/gambol99/go-marathon
-  version: 15ea23e360abb8b25071e677aed344f31838e403


Could you add/update only your dependencies?

Not very sure why this shows up as a diff, the glide.yaml has d672c6fbb499596869d95146a26e7d0746c06c54 version for github.com/gambol99/go-marathon which came from a prior commit (https://github.com/containous/traefik/blame/master/glide.yaml#L96)

It's not only for go-marathon. It's a side affect of the glide get.
You must use a workaround:

Revert the glide manifest and lock files.

Use the github.com/multiplay/glide-pin plugin to pin all versions in the manifest file to the ones given in the lock file.

Add your dependencies glide get

Run glide update -v to update the lock file.

Undo the pinning.

Add your dependencies in glide.yml

Use glide-hash to get the manifest file hash and put it in the lock file to please validate-glide.

Look at #1332 (comment)

Tried to follow those steps, only to run into errors :(

Invoked glide pin followed by glide get github.com/go-kit/kit/metrics/dogstatsd commands and things exploded:

[ERROR] Error scanning github.com/docker/docker/cli/config: open /Users/aantonov/.glide/cache/src/https-github.com-docker-docker/cli/config: no such file or directory [ERROR] This error means the referenced package was not found. [ERROR] Missing file or directory errors usually occur when multiple packages [ERROR] share a common dependency and the first reference encountered by the scanner [ERROR] sets the version to one that does not contain a subpackage needed required [ERROR] by another package that uses the shared dependency. Try setting a [ERROR] version in your glide.yaml that works for all packages that share this [ERROR] dependency. [ERROR] Error scanning github.com/docker/docker/cli/config/configfile: open /Users/aantonov/.glide/cache/src/https-github.com-docker-docker/cli/config/configfile: no such file or directory [ERROR] This error means the referenced package was not found. [ERROR] Missing file or directory errors usually occur when multiple packages [ERROR] share a common dependency and the first reference encountered by the scanner [ERROR] sets the version to one that does not contain a subpackage needed required [ERROR] by another package that uses the shared dependency. Try setting a [ERROR] version in your glide.yaml that works for all packages that share this [ERROR] dependency. [INFO] --> Fetching updates for github.com/xeipuuv/gojsonpointer. [ERROR] Error scanning github.com/docker/docker/cli/command/image/build: open /Users/aantonov/.glide/cache/src/https-github.com-docker-docker/cli/command/image/build: no such file or directory [ERROR] This error means the referenced package was not found. [ERROR] Missing file or directory errors usually occur when multiple packages [ERROR] share a common dependency and the first reference encountered by the scanner [ERROR] sets the version to one that does not contain a subpackage needed required [ERROR] by another package that uses the shared dependency. Try setting a [ERROR] version in your glide.yaml that works for all packages that share this [ERROR] dependency. [INFO] --> Fetching updates for golang.org/x/time. [INFO] --> Setting version for golang.org/x/time to a4bde12657593d5e90d0533a3e4fd95e635124cb. [INFO] --> Fetching updates for github.com/Azure/go-ansiterm. [INFO] --> Fetching updates for github.com/gorilla/mux. [INFO] --> Setting version for github.com/gorilla/mux to e444e69cbd2e2e3e0749a2f3c717cec491552bbf. [INFO] --> Fetching updates for github.com/vbatts/tar-split. [INFO] --> Fetching updates for github.com/opencontainers/runtime-spec. [ERROR] Failed to retrieve a list of test dependencies: Error resolving imports

Can you try two things:

Run glide cc first thing. (This will clear the glide cache which is sometimes known to cause problems.)

use script/glide.sh <command> instead of glide <command> everywhere.

In that case, it becomes more and more likely that we are really running into the problem the glide error describes: conflicting dependencies.

I can double check on my end. Could you tell me which additional dependencies the PR needs?

I've been trying to add:

github.com/go-kit/kit/metrics/dogstatsd github.com/go-kit/kit/metrics/statsd github.com/go-kit/kit/metrics/multi github.com/go-kit/kit/log github.com/go-kit/kit/util/conn github.com/stvp/go-udp-testing

Would greatly appreciate if you could see if it will work on your end.

I'm afraid I ran into similar problem: Trying to upgrade github.com/stvp/go-udp-testing alone yields many conflicts with the client-go library.

I'm having a hard time understanding why we run into these conflicts since go-udp-testing looks super simple and seemingly doesn't have any direct or transitive dependencies with go-client.

Maybe we're holding it wrong at some point...

@aantono could you squash all commits and remove changes on glide.yml, glide.lock and vendor? I will try to resolve the problem after that.

ldez · 2017-06-09T09:55:41Z

middlewares/metrics.go

+		histograms = append(histograms, m.getReqDurationHistogram())
+		retryCounters = append(retryCounters, m.getRetryCounter())
+	}
+	var mm MultiMetrics


move this line after the next empty line.

ldez · 2017-06-09T09:55:56Z

middlewares/statsd_test.go

+	udp.ShouldReceiveAll(t, expected, func() {
+		n.ServeHTTP(recorder, req1)
+		n.ServeHTTP(recorder, req2)
+		//		body := recorder.Body.String()


Dead code ?

aantono · 2017-06-12T18:56:24Z

Rebased against master, removing changes on glide.yml, glide.lock and vendor. @ldez, this is all yours! ;)

ldez · 2017-06-12T22:24:37Z

@aantono it's ready. 🎉

ldez

You should use safe.Go instead of the classic go routines.

ldez · 2017-06-12T22:26:34Z

middlewares/datadog.go

+
+		report := time.NewTicker(pushInterval)
+
+		go datadogClient.SendLoop(report.C, "udp", address)


safe.Go(func() { datadogClient.SendLoop(report.C, "udp", address) })

ldez · 2017-06-12T22:29:17Z

middlewares/statsd.go

+
+		report := time.NewTicker(pushInterval)
+
+		go statsdClient.SendLoop(report.C, "udp", address)


safe.Go(func() { statsdClient.SendLoop(report.C, "udp", address) })

aantono · 2017-07-18T16:10:18Z

No worries, You bet! Looks like the glide.lock is again in conflict... guess its time for @ldez to do his magic again! 😉

timoreimann · 2017-07-19T21:53:17Z

middlewares/datadog.go

+	if datadogTicker == nil {
+		address := config.Address
+		if len(address) == 0 {
+			address = "localhost:8125"


Why do we default to this address? Is this somehow conventional in Datadog?

Instinctively, I tend to go with requiring proper configuration and -- if missing -- return an error, unless a default makes sense.

The default DataDog statsD listener is running on localhost:8125, which is a well known port, so it is very common for people to not define it explicitly and use the default (be that in DataDog agent config, or in the clients) (See default dd-agent config for reference - https://github.com/DataDog/dd-agent/blob/master/datadog.conf.example#L163)

timoreimann · 2017-07-19T21:54:08Z

middlewares/datadog.go

+		}
+		pushInterval, err := time.ParseDuration(config.PushInterval)
+		if err != nil {
+			log.Warnf("Unable to parse %s into pushInterval", config.PushInterval)


I think the log message should mention that we'll be falling back to a default value.

timoreimann · 2017-07-19T22:02:49Z

middlewares/datadog.go

+
+// InitDatadogClient initializes metrics pusher and creates a datadogClient if not created already
+func InitDatadogClient(config *types.Datadog) *time.Ticker {
+	if datadogTicker == nil {


I suppose we do not enter this function from multiple goroutines concurrently, and hence don't need to synchronize access?

Correct. We only call InitDatadogClient once when the server starts.

An alternative to comparing the ticker could then be to use sync.Once. This may also help us avoid accidentally initializing the client more than once if for whatever reason we toggle the ticker value inadvertently (say, due to a bug).

More of a suggestion though.

sync.Once sounds like a good idea. I've tried to add it, but realized that I still need to keep track of the ticker reference, so that I could stop it once the shutdown occurs. Since I have to keep the ticker variable around anyways, do you want to introduce another variable to store the once, or just let it be and use the null check instead?

I think it's good enough in this case. Thanks for giving it a shot. 👍

timoreimann · 2017-07-19T22:12:08Z

middlewares/statsd.go

+}
+
+// InitStatsdClient initializes metrics pusher and creates a statsdClient if not created already
+func InitStatsdClient(config *types.Statsd) *time.Ticker {


The entire file, but especially this function look very similar to what is implemented in datadog.go. Is there any chance to factor out the duplicated code?

It's two different features, factorize don't seems a good idea.
For me, it's like consul.go, etcd.go or boltdb.go.

While I thought about trying to generalize and factorize the code out, like @ldez mentioned, while the 2 files look very similar in pattern, they internally use different go-kit/metrics structures to initialize metrics, name them and return outbound structures, so there isn't much that can actually be shared, besides the logical steps of operation, which are similar due to the similarity in DataDog being an enhancement on top of StatsD protocol.

@aantono The initial function is what I was mostly after. They look pretty much identical except for the ticker they reference and the receiver they trigger the loop on. That should be fairly easy to parameterize.

More importantly to me, refactoring should allow us to squash two fairly big tests into a single one.

When you say

besides the logical steps of operation, which are similar due to the similarity in DataDog being an enhancement on top of StatsD protocol.
(emphasis by me)

to me that's really a signal for how the two share certain attributes (i.e., the function in question) that we could reflect code-wise by a central implementation.

So that's my view. Maybe @emilevauge can step in and become the tipping point. I'll be happy to follow whatever majority constitutes. :-)

@ldez tbh, seeing three k/v store implementations spanning 30+ LOC each that really only differ by a single store type they pass into a function is rather an anti-example to me. 😮

So I looked once again at the InitDataDogClient and InitStatsDClient code and the only part they share is this:

address := config.Address if len(address) == 0 { address = "localhost:8125" } pushInterval, err := time.ParseDuration(config.PushInterval) if err != nil { log.Warnf("Unable to parse %s into pushInterval", config.PushInterval) pushInterval = 10 * time.Second }

Given that it looks like boiler-plate code, it's really just a validation of config properties (but keep in mind that config is of different struct type: types.Datadog vs. types.Statsd), they just happen to both have address and pushInterval properties).
The rest of the function handles initializing different clients and each provider has their own ticker as well, which is a statically managed variable, so not that great for factorizing. IMHO it would be more code and will make it harder to read and reason about the code if we to try to come up with a base config type (not even sure what to call it) and then have the 2 concrete types encapsulate it. While I do agree that on the surface this does seem like a good possible case for boiler-plate reduction, once you look deeper, that does not turn out to be so, but curious to hear what @emilevauge thinks about this.

Alright, you convinced me: I failed to see that the config structs are of different types as well. With that in mind, extracting anything doesn't seem too useful.

Appreciate the technical analysis to solve the matter. 👏

timoreimann · 2017-07-19T22:12:39Z

middlewares/statsd_test.go

@@ -0,0 +1,52 @@
+package middlewares


Tests look very similar to datadog_test.go too. Reuse opportunity?

same thing here, it's two different features.

timoreimann · 2017-07-19T22:14:11Z

server/server.go

+		}
+		if globalConfig.Web.Metrics.Datadog != nil {
+			metric := middlewares.NewDataDog(name)
+			log.Debugf("Configured DataDog Metrics pushing to %s once every %s", globalConfig.Web.Metrics.Datadog.Address, globalConfig.Web.Metrics.Datadog.PushInterval)


s/Metrics/metrics/

just to clarify, are you asking to replace "DataDog Metrics" part of the string in the log.Debugf with DataDog metrics lowercased metrics? If that's the case, I would argue that DataDog Metrics is a name-proper of the monitoring provider (just like Prometheus Metrics before it, but if you insist, would be happy to change it.

Ah sorry, I think I was being too complicated: Yes, I just meant to use lower case in the log message.

I'm only familiar with Prometheus where I'm not aware of a proper name like Prometheus Metrics. If you could point me at any references (e.g., docs) showing this, I'll be happy to rest my case.

timoreimann · 2017-07-19T22:14:15Z

server/server.go

-		return metrics
+		if globalConfig.Web.Metrics.StatsD != nil {
+			metric := middlewares.NewStatsD(name)
+			log.Debugf("Configured StatsD Metrics pushing to %s once every %s", globalConfig.Web.Metrics.StatsD.Address, globalConfig.Web.Metrics.StatsD.PushInterval)


s/Metrics/metrics/

timoreimann · 2017-07-19T22:17:38Z

server/server.go

+	metricsEnabled := globalConfig.Web != nil && globalConfig.Web.Metrics != nil
+	if metricsEnabled {
+		if globalConfig.Web.Metrics.Datadog != nil {
+			middlewares.StopDatadogClient()


Do we actually need all those ifs before? It looks like StopDatadogClient won't try to stop anything anyway unless the ticker was previously set (i.e., we have configured Datadog for sure).

We probably don't, at least not at the current moment, I just put them in there for security measure, in case later on the internals of StopDAtadogClient() change in a way that would require such a check to be done, but we could forget to do it. I would be happy to remove the if checks, if needed.

I'd argue that StopDatadogClient should decide if the client still needs to be stopped or not, rather than relying on static configuration parameters. The function already does this by means of using the ticker value as a flag, so overall I think we can remove the if blocks.

timoreimann · 2017-07-19T22:17:48Z

server/server.go

+			middlewares.StopDatadogClient()
+		}
+		if globalConfig.Web.Metrics.StatsD != nil {
+			middlewares.StopStatsdClient()


Same question about the if'ing here.

timoreimann · 2017-07-19T22:21:54Z

I think we're also missing a docs update somewhere to inform our users about the possibility to use Datadog or Statsd for metrics collection.

timoreimann · 2017-07-20T00:20:03Z

middlewares/statsd.go

+	if statsdTicker == nil {
+		address := config.Address
+		if len(address) == 0 {
+			address = "localhost:8125"


Shouldn't this use port 9125 given what you said here about the default ports?

My mistake, I confused StatsD with statsd-exporter project, which defaults to 9125. The default port for both StatsD and dogStatsD is indeed 8125, since the protocols are largely compatible (meaning you can send plain StatsD format data to dogStatsD agent, and it will accept it)

timoreimann

LGTM. Ready to merge once that dreaded merge conflict on the glide lock file (eek!) is resolved. :)

aantono · 2017-07-20T15:52:09Z

Sounds good... Looks like this is another opportunity for @ldez to shine and show off his mad "Glide" skills! 😜

…e redundant.

mjeri

Sorry to comment on the closed PR. Didn't know a better place to raise my concerns. @aantono can you have a look at the comments and if necessary make a clean up PR?

Especially the point regarding DataDog/StatsD+Retries is concerning to me.

mjeri · 2017-07-21T11:23:55Z

middlewares/datadog.go

+	"github.com/go-kit/kit/metrics/dogstatsd"
+)
+
+var _ Metrics = (Metrics)(nil)


Does this statement have any effect? I saw these kind of constructs to verify a concrete implementation is implementing an interface. To have the Metrics on the right hand side doesn't make any sense to me. Can you explain it or clean it up?

The right hand side should have been (*DataDog)(nil). Fixed.

mjeri · 2017-07-21T11:25:26Z

middlewares/datadog.go

+	return dd.reqDurationHistogram
+}
+
+func (dd *Datadog) getRetryCounter() metrics.Counter {


As retryCounter never gets initialised won't this lead to a panic when having DataDog implementation (same for StatsD) + Retries activated? Did you try this out?

@marco-jantke Great catch... the retry metric was added after the PR was already open, so I must've lost the change during one of the numerous rebases. The good news is that the code won't panic, since there is a m.retryMetrics != nil check before the counter is used, but the bad news is that we are missing retry metrics for DataDog & StatsD. Thanks a lot for noticing, I'll add those to the initialization block and open another PR.

mjeri · 2017-07-21T11:25:44Z

middlewares/metrics.go

+	return mm.retryCounter
+}
+
+func (mm *MultiMetrics) getWrappedMetrics() *[]Metrics {


AFAICS this method is used nowhere.

Yep, you are correct... Same PR that added retryMetrics, removed the getWrappedMetrics() method. It was originally present on the initial Metrics interface. Will clean-up

mjeri · 2017-07-21T11:26:27Z

middlewares/statsd.go

+// and the number of requests partitioned by status code and method.
+// - number of requests partitioned by status code and method
+// - request durations
+// - amount of retries happened


It does not expose amount of retries happened AFAICS. Same for DataDog implementation.

This merge removes the integration of DataDog and StatsD that was introduced with this PR: traefik#1701 This integration is highly incompatible with our current Prometheus metrics implementation that we have in dustify. Therefore everything specific to those metrics integration was removed, except the glide adaption. They don't hurt for now and better not to start getting merge conflicts in them in there.

ldez added area/middleware/metrics kind/enhancement a new or improved feature. status/1-needs-design-review contributor/needs-resolve-conflicts labels Jun 2, 2017

ldez requested a review from emilevauge June 2, 2017 11:44

aantono force-pushed the Issue834 branch 7 times, most recently from 3de1539 to 01b45ba Compare June 6, 2017 03:23

emilevauge added status/2-needs-review and removed status/1-needs-design-review labels Jun 6, 2017

aantono force-pushed the Issue834 branch from 01b45ba to 7fe05d3 Compare June 7, 2017 16:31

aantono force-pushed the Issue834 branch from 7fe05d3 to 0735404 Compare June 7, 2017 16:48

ldez removed the contributor/needs-resolve-conflicts label Jun 7, 2017

ldez requested changes Jun 9, 2017

View reviewed changes

aantono force-pushed the Issue834 branch 5 times, most recently from 51a26e7 to 143dc17 Compare June 12, 2017 18:38

ldez force-pushed the Issue834 branch 3 times, most recently from 4741f30 to c5d50d4 Compare June 12, 2017 21:52

ldez requested changes Jun 12, 2017

View reviewed changes

ldez removed the contributor/waiting-for-corrections label Jul 18, 2017

ldez force-pushed the Issue834 branch from d270ee0 to ba2483f Compare July 19, 2017 08:28

timoreimann suggested changes Jul 19, 2017

View reviewed changes

ldez added the contributor/waiting-for-corrections label Jul 19, 2017

timoreimann reviewed Jul 20, 2017

View reviewed changes

timoreimann approved these changes Jul 20, 2017

View reviewed changes

ldez added status/3-needs-merge and removed contributor/waiting-for-corrections status/2-needs-review labels Jul 20, 2017

ldez force-pushed the Issue834 branch from 874f13a to ae78d89 Compare July 20, 2017 22:04

aantono and others added 9 commits July 21, 2017 00:05

Added support for DataDog and StatsD monitoring

caba277

Fixed goroutine spawns

fc6220a

Moved DD/StatsD client initialization into a startHTTPServers()

bbbb7e4

Added DD/SD client stop calls when 'server.Close()' is executed

810a5f1

refactor: fix comment on exported Stop methods.

9a646a7

Removed checks before stopping DataDog and StatsD clients, as they ar…

e864a38

…e redundant.

Added documentation

225a1c7

Cleanup of logging messages to make them more descriptive/accurate

613c270

chore(glide): Add dependencies.

88c99fd

ldez force-pushed the Issue834 branch from ae78d89 to 88c99fd Compare July 20, 2017 22:06

ldez removed the status/3-needs-merge label Jul 20, 2017

ldez merged commit 69c628b into traefik:master Jul 20, 2017

mjeri reviewed Jul 21, 2017

View reviewed changes

aantono mentioned this pull request Jul 21, 2017

Added RetryMetrics to DataDog and StatsD providers #1884

Merged

aantono mentioned this pull request Aug 21, 2017

Support multi-port service routing for containers running on Marathon #1621

Closed

ldez mentioned this pull request Aug 29, 2017

StatsD (e.g. DataDog) support #834

Closed


		report := time.NewTicker(pushInterval)

		go datadogClient.SendLoop(report.C, "udp", address)


		report := time.NewTicker(pushInterval)

		go statsdClient.SendLoop(report.C, "udp", address)

DataDog and StatsD Metrics Support #1701

DataDog and StatsD Metrics Support #1701

Conversation

aantono commented Jun 2, 2017

aantono commented Jun 7, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ldez Jun 9, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aantono commented Jun 12, 2017

ldez commented Jun 12, 2017

ldez left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aantono commented Jul 18, 2017 • edited

Choose a reason for hiding this comment

aantono Jul 19, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

timoreimann Jul 20, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ldez Jul 19, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aantono Jul 20, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ldez Jul 19, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aantono Jul 20, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

timoreimann commented Jul 19, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

timoreimann left a comment

Choose a reason for hiding this comment

aantono commented Jul 20, 2017

mjeri left a comment • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ldez Jun 9, 2017 •

edited

aantono commented Jul 18, 2017 •

edited

aantono Jul 19, 2017 •

edited

timoreimann Jul 20, 2017 •

edited

ldez Jul 19, 2017 •

edited

aantono Jul 20, 2017 •

edited

ldez Jul 19, 2017 •

edited

aantono Jul 20, 2017 •

edited

mjeri left a comment •

edited