I created a test endpoint in my webapp that would call itself in a loop via Java httpClient. Using another tool, I can see these requests are completing successfully. Consider this a poor man's load test.
Dynatrace does not appear to be capturing this request back to the application. All other transactions are captured as far as I can tell. Is this possibly causing a bug in the transaction parsing? These requests are also not showing up on "dynamic requests" on the throughput timeline as I have all traffic stopped other than these tests.
In addition, I don't even see the "parent" request that is responsible for kicking off the GETs back to self.
Can you give us some more information Chesea? Could you link your SaaS or Managed tenant where you are experiencing this?
And I'm a bit confused. You say "...that would call itself in a loop via Java httpClient" then say that you don't see the parent request. If this request originates within it's own service then you wouldn't expect to see anything before that service, if that makes sense.
@hayden Here is my setup, dynatrace is watching all the services involved in this request except for the database.
Primary application receives web request (endpoint A)
Which makes a web request in a loop back to the primary application (endpoint B) using httpClient
EndPoint B calls a separate service, Endpoint, C via httpClient
Which calls ElasticSearch (This instance of elasticsearch is intentionally shut down)
I found out that if I lowered the number of requests endpoint A makes of endpoint B from 200 to something much smaller such as 5, the transactions were showing up in the dynamic request pure path view-- the main parent request as well as the child web requests (transactions). Maybe it's too much for Dynatrace to track in this scenario.
In addition, I'm on to my next problem which is figuring out why dynatrace doesn't sound the alarm bells that the secondary application is trying to hit a REST based API that is unavailable. This is all our organization wants -- an intelligent tool that can get to root cause quickly. This evaluation was going so well until I started getting into the weeds trying to simulate problems.
Hm, that sounds interesting and like something else is going on behind the scenes. Dynatrace should have no problem handling that type of load - that's what it was designed for!
The second problem - is the REST API that is unavailable (elastic search) sending back any sort of response code that or timeout happening that causes a failure? It could be that the requests are going there and coming back with a status code that isn't represented as a failure. I'm not sure what else could be going on in that scenario.
Overall, I recommend opening a support ticket for our teams to dig in and take a further look at your environment. You can do that here: https://www.dynatrace.com/support/. Click on contact support, and open a support ticket for Dynatrace.
Hope this helps get you to a solution,